Unsupervised Active Visual Search With Monte Carlo Planning Under Uncertain Detections
Open accessThe DOI matches its Crossref record; 62 of the 62 references it cites were found in a registry or an index; 4 of the 17 sources its claims cite were checked for support.
Abstract
We propose a solution for Active Visual Search of objects in an environment, whose 2D floor map is the only known information. Our solution has three key features that make it more plausible and robust to detector failures compared to state-of-the-art methods: i) it is unsupervised as it does not need any training sessions. ii) During the exploration, a probability distribution on the 2D floor map is updated according to an intuitive mechanism, while an improved belief update increases the effectiveness of the agent's exploration. iii) We incorporate the awareness that an object detector may fail into the aforementioned probability modelling by exploiting the success statistics of a specific detector. Our solution is dubbed POMP-BE-PD (Pomcp-based Online Motion Planning with Belief by Exploration and Probabilistic Detection). It uses the current pose of an agent and an RGB-D observation to learn an optimal search policy, exploiting a POMDP solved by a Monte-Carlo planning approach. On the Active Vision Dataset Benchmark, we increase the average success rate over all the environments by a significant 35 % while decreasing the average path length by 4 % with respect to competing methods. Thus, our results are state-of-the-art, even without any training procedure.
Overview
CoChat Summary
TL;DR
POMP-BE-PD is an unsupervised active visual search method that plans online with POMCP in a known 2D floor map, using probabilistic object detection and an improved belief update. On the Active Vision Database benchmark it reports higher success rates than competing methods and its prior POMP formulation without any training procedure.
Key findings
- With the real detector, POMP-BE-PD reaches average SR 0.54, APL 39.44, and SPL 0.47 across the tested AVDB scenes, versus POMP’s 0.40, 41.13, and 0.50; the paper reports this as a 35% relative increase in success rate and a 4% decrease in average path length, or a 14% absolute SR increase in the conclusion.
- Probabilistic detection reduces false positives by about 32%, and the docking strategy reduces docking error by about 35.7% (Fig. 8).
- With ground-truth annotations, the improved belief update alone (POMP-BE) raises medium-scene SR from 0.76 to 0.85 and lowers APL from 29.64 to 25.50; hard-scene SR rises from 0.48 to 0.70.
- On the GT-object protocol in Table 1, POMP-BE-PD achieves average SR 0.90 versus POMP’s 0.76, with APL 18.10 versus 17.07 and SPL 0.72 versus 0.75.
Methods
- The task is modeled as a POMDP solved by Partially Observable Monte Carlo Planning (POMCP); the agent uses its pose and RGB-D observations, with a pose graph, candidate object locations, and a visibility matrix derived from the known 2D floor map.
- Detection uses the Target Driven Instance Detector (TDID), keeping detections with score > 0.9; class-specific F1 statistics feed a probabilistic detection vector and Bayesian update over object locations, with exploration terminating when a visible location’s probability exceeds threshold τ.
- Experiments use the Active Vision Database benchmark with 8 scenes (2 easy, 3 medium, 3 hard), metrics SR, APL, and SPL, a 200-step limit, and comparisons against Random Walk, EAT, DQN, DQN-TAM, and POMP.
Limitations
- Stated: no official code for published comparison methods was available, so comparisons follow the protocol from [9], which uses only GT annotations and a limited number of scenes; DQN and DQN-TAM use only two scenes and a different object subset, making their average column not meaningful for fair comparison.
- Evident from the text: the method assumes a known 2D floor map and relies on an off-the-shelf object detector whose class-specific F1 statistics must be available; performance depends on threshold τ and simulation budget.
- Evident from the text: evaluation is limited to the AVDB benchmark and selected scenes, and probabilistic detection can increase APL and reduce SPL in some settings.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 4 Not valid · 7 Not checked · 31 Own results and interpretations
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“It is worth noting that for achieving these results both DQN approaches require 13 scenarios for training the best policy, while our method requires no training at all.”
§ 4 EXPERIMENTS > 4.1 Quantitative results
What the sources say
- Could not verifycites [47] matched to “Learning State-Variable Relationships in POMCP: A Framework for Mobile Robots” (2022) sourceExperimental — accuracy not yet measured
The Introduction discusses deep reinforcement learning learning policies from observations, but does not describe two DQN approaches or their training requirements. Searches of our copy for DQN and 13 scenarios found no passage establishing that either approach required 13 scenarios to train its best policy; the cited paper’s Experiments section instead discusses MRF learning. (LLM, conf=0.76)
- Not checked — no full text of the sourcecites [48] matched to “Review - Mastering the game of Go with deep neural networks and tree search” (2017) source
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In all the scenarios, POMP-PD increments the SR by a large margin (19% - 25%) over our previous formulation POMP.”
§ 4 EXPERIMENTS > 4.2 Ablation studies
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“After every step in the real world, a Bayesian inference, combined with a probability distribution over all possible object locations, allows us to reduce the false positives error by 32%.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“This strategy achieves better performance than the Robust Visual Docking introduced in [12].”
§ 3 METHOD > 3.3 Exploration, localisation and approach
What the sources say
- Not matchedcites [12] Silver, Monte-Carlo planning in large POMDPs
Its entry has no title and year that we can search by.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
- Not matched
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“First of all, we can notice a significant reduction (∼32% decrease) of false positives when introducing the Probabilistic Detection approach, thus increasing the robustness of our method.”
§ 4 EXPERIMENTS > 4.2 Ablation studies
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
14 more claims
Create a free account to see all 42 claims and every source verdict.
- It may state two separate claims as one. 19
- Probably not a substantial claim. 5
- The quote only partly states this claim. 2
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 3 METHOD > 3.3 Exploration, localisation and approach
“The new belief is sampled from a uniform distribution over locations satisfying the pp constraint –namely, having the searched object in positions belonging to H \ pp.”
§ 3 METHOD > 3.3 Exploration, localisation and approach
“An auxiliary variable pp stores the current list of object locations not been observed yet, where pp = {j ∈ H | j not yet observed}.”
Citations
Cites
A note on two problems in connexion with graphs
Dijkstra, E. W. (1959) · Numerische Mathematik · doi:10.1007/bf01386390 · Cited by 29,065
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612
(2006) · Springer eBooks · Cited by 8,123
Planning and acting in partially observable stochastic domains
Kaelbling, L. P., Littman, M. L., & Cassandra, A. R. (1998) · Artificial Intelligence · doi:10.1016/s0004-3702(98)00023-x · Cited by 4,149 · cited 2×
“We formulate the AVS problem as a Partially Observable Markov Decision Process (POMDP), which is a standard framework for modeling sequential decision processes under uncertainty in dynamical environments [26].”
BackgroundMethodA general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. (2018) · Science · doi:10.1126/science.aar6404 · Cited by 3,562 · cited 1×
“Popular MCTS-based approaches have been recently used also for developing agents with superhuman performance in the game of Go [43], [44].”
MethodA Survey of Monte Carlo Tree Search Methods
Browne, C. B., Powley, E., Whitehouse, D., Lucas, S. M., Cowling, P. I., Rohlfshagen, P., Tavener, S., Perez, D., Samothrakis, S., & Colton, S. (2012) · IEEE Transactions on Computational Intelligence and AI in Games · doi:10.1109/tciaig.2012.2186810 · Cited by 2,970 · cited 1×
“One of the most recent and efficient strategies for solving POMDPs in an approximate way is Monte Carlo Tree Search (MCTS) [28], [29], [30].”
MethodBandit Based Monte-Carlo Planning
Kocsis, L., & Szepesvári, C. (2006) · Lecture Notes in Computer Science · doi:10.1007/11871842_29 · Cited by 2,855 · cited 3×
“At each step, an action a is selected using UCT when the current history is inside the tree, and a uniform policy when the current history is outside the tree.”
MethodMonte-Carlo Planning in Large POMDPs
Silver, D., & Veness, J. (2010) · DSpace@MIT (Massachusetts Institute of Technology) · Cited by 1,408
A Note on Two Problems in Connexion with Graphs
Dijkstra, E. W. (2022) · Edsger Wybe Dijkstra · doi:10.1145/3544585.3544600 · Cited by 1,353
The Complexity of Markov Decision Processes
Papadimitriou, C. H., & Tsitsiklis, J. N. (1987) · Mathematics of Operations Research · doi:10.1287/moor.12.3.441 · Cited by 1,329 · cited 1×
“Computing exact solutions for non-trivial POMDPs is often computationally intractable [27], but in the recent years impressive progress was made to develop approximate solvers.”
BackgroundEfficient Selectivity and Backup Operators in Monte-Carlo Tree Search
Coulom, R. (2007) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-75538-8_7 · Cited by 1,240 · cited 1×
“One of the most recent and efficient strategies for solving POMDPs in an approximate way is Monte Carlo Tree Search (MCTS) [27], [28], [29].”
MethodMastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm
Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. (2017) · arXiv · doi:10.48550/arxiv.1712.01815 · Cited by 1,102
Review - Mastering the game of Go with deep neural networks and tree search
Babbar, S. (2017) · Unpublished · doi:10.13140/rg.2.2.18893.74727 · Cited by 929
On evaluation of embodied navigation agents
Anderson, P., Chang, A., Chaplot, D. S., Dosovitskiy, A., Gupta, S., Koltun, V., Kosecka, J., Malik, J., Mottaghi, R., Savva, M., et al. (2018) · arXiv · doi:10.48550/arxiv.1807.06757 · Cited by 622 · cited 2×
“We consider three metrics: Success Rate (SR) [56] is considered the main metric of this work, and it is defined as the percentage of times the agent successfully reaches one of the destination poses (as provided in AVDB) over the total number of trials (a larger value indicates a more effective…”
BackgroundMethodLearning to Navigate in Complex Environments
Mirowski, P., Pascanu, R., Viola, F., Soyer, H., Ballard, A. J., Banino, A., Denil, M., Goroshin, R., Sifre, L., Kavukcuoglu, K., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1611.03673 · Cited by 323
Object Goal Navigation using Goal-Oriented Semantic Exploration
Chaplot, D. S., Gandhi, D., Gupta, A., & Salakhutdinov, R. (2020) · arXiv · doi:10.48550/arxiv.2007.00643 · Cited by 321
Thrun, S. (1999) · neural information processing systems · Cited by 303
Visual Representations for Semantic Target Driven Navigation
Mousavian, A., Toshev, A., Fiser, M., Kosecka, J., Wahid, A., & Davidson, J. (2019) · 2019 International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2019.8793493 · Cited by 224 · cited 1×
“This limitation is shared with other approaches that learn optimal policies using Asynchronous Advantage Actor Critic (A3C) algorithm [18], Long Short Term Memories (LSTM) architectures [19], and Transformer networks coupled with deep Q-Learning [20].”
MethodA dataset for developing and benchmarking active vision
Ammirato, P., Poirson, P., Park, E., Kosecka, J., & Berg, A. C. (2017) · 2017 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2017.7989164 · Cited by 187 · cited 4×
“With respect to the standard protocol defined in the benchmark paper [9], this protocol provides results only using GT annotations for object detection, and on a limited number of scenes.”
BackgroundMethodInfluentialScene Memory Transformer for Embodied Agents in Long-Horizon Tasks
Fang, K., Toshev, A., Fei-Fei, L., & Savarese, S. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00063 · Cited by 178 · cited 1×
“This limitation is shared with other approaches that learn optimal policies using Asynchronous Advantage Actor Critic (A3C) algorithm [18], Long Short Term Memories (LSTM) architectures [19], and Transformer networks coupled with deep Q-Learning [20].”
MethodObjectNav Revisited: On Evaluation of Embodied Agents Navigating to Objects
Batra, D., Gokaslan, A., Kembhavi, A., Maksymets, O., Mottaghi, R., Savva, M., Toshev, A., & Wijmans, E. (2020) · arXiv · doi:10.48550/arxiv.2006.13171 · Cited by 177 · cited 1×
“Within this field, AVS is a task focused on detecting and approaching a specific object [15].”
BackgroundOn the relative complexity of active vs. passive visual search
Tsotsos, J. K. (1992) · International Journal of Computer Vision · doi:10.1007/bf00128132 · Cited by 164 · cited 1×
“A MONG the most interesting areas of robotics is the problem of Active Visual Search (AVS) [1], in which an intelligent robotic agent must autonomously find an object located far from it, moving and exploring its surroundings through egocentric visual sensors.”
BackgroundPONI: Potential Functions for ObjectGoal Navigation with Interaction-free Learning
Ramakrishnan, S. K., Chaplot, D. S., Al-Halah, Z., Malik, J., & Grauman, K. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01832 · Cited by 154 · cited 1×
“On top of such semantic exploration, the perception skills in terms of where to look and the navigation can be further disentangled [23] for an improved success rate.”
BackgroundMotion Planning for Carlike Robots Using a Probabilistic Learning Approach
Svestka, P., & Overmars, M. H. (1997) · The International Journal of Robotics Research · doi:10.1177/027836499701600201 · Cited by 103 · cited 1×
“In [42],theauthorspresented aprobabilistictechniqueforsolvingthelearningmotionplan-ningprobleminstaticenvironments.”
Planning for robotic exploration based on forward simulation
Lauri, M., & Ritala, R. (2016) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2016.06.008 · Cited by 98 · cited 1×
“A fewofthemarerelatedtotheexplorationofpartiallyknown environments[44]andthefind-and-followofpeople[45]with robots.”
Habitat-Matterport 3D Semantics Dataset
Yadav, K., Ramrakhya, R., Ramakrishnan, S. K., Gervet, T., Turner, J., Gokaslan, A., Maestre, N., Chang, A. X., Batra, D., Savva, M., et al. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.00477 · Cited by 92 · cited 1×
“…at a random starting pose in an unknown environment to find a given instance of an object category using only sensory inputs to navigate, where large-scale datasets of 3D real-world spaces with densely annotated semantics are also made available to facilitate training and testing the models [22].”
BackgroundActive Object Detection With Multistep Action Prediction Using Deep Q-Network
Han, X., Liu, H., Sun, F., & Zhang, X. (2019) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2019.2890849 · Cited by 86 · cited 2×
“AVS with deep learning is viable using Deep Reinforcement Learning techniques [10], [11], [17], where visual neural embed-dingsareoftenexploitedforactionpolicytraining.”
MethodScalable Planning and Learning for Multiagent POMDPs
Amato, C., & Oliehoek, F. (2015) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v29i1.9439 · Cited by 79 · cited 1×
“A version of POMCP for scalable planning in multiagent POMDPs is presented in [33], which it introduced model learning in POMDPs considering also the uncertainty about model parameters in the belief.”
MethodMonte-Carlo Tree Search for Constrained POMDPs
Lee, J., Kim, G.-H., Poupart, P., & Kim, K.-E. (2018) · neural information processing systems · Cited by 79
Learning in POMDPs with Monte Carlo Tree SearchPreprint
Katt, S., Oliehoek, F. A., & Amato, C. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1806.05631 · Cited by 77
Probabilistic motion planning among moving obstacles following typical motion patterns
Fulgenzi, C., Spalanzani, A., & Laugier, C. (2009) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2009.5354755 · Cited by 71
Using Qualitative Spatial Relations for indirect object search
Kunze, L., Doreswamy, K. K., & Hawes, N. (2014) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2014.6906604 · Cited by 56 · cited 1×
“A probabilistic approach is proposed in [16], where the likelihood of the target increases when objects which are expected to be co-occurring are detected.”
BackgroundBayesian Reinforcement Learning in Factored POMDPs
Katt, S., Oliehoek, F. A., & Amato, C. (2019) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/msbg2685 · Cited by 43
Topological spatial relations for active visual search
Sjöö, K., Aydemir, A., & Jensfelt, P. (2012) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2012.06.001 · Cited by 41
Attention-based active visual search for mobile robots
Rasouli, A., Lanillos, P., Cheng, G., & Tsotsos, J. K. (2019) · Autonomous Robots · doi:10.1007/s10514-019-09882-z · Cited by 41 · cited 1×
“…2D grid map of the scene in our POMCP modelling: blue dots are the possible object locations, purple crosses are the possible robot poses. contexts, such as the domotic field [2], [3], [4], personal assistance [5], search and rescue [6], [7], and the very intriguing Mars exploration [8].”
BackgroundContinuous real time POMCP to find-and-follow people by a humanoid service robot
Goldhoorn, A., Garrell, A., Alquézar, R., & Sanfeliu, A. (2014) · IEEE-RAS International Conference on Humanoid Robots (Humanoids) · doi:10.1109/humanoids.2014.7041445 · Cited by 37 · cited 1×
“A fewofthemarerelatedtotheexplorationofpartiallyknown environments[44]andthefind-and-followofpeople[45]with robots.”
Target Driven Instance DetectionPreprint
Ammirato, P., Fu, C.-Y., Shvets, M., Košecká, J., & Berg, A. C. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1803.04610 · Cited by 30 · cited 4×
“2) Probabilistic Detection: We equipped our agent with the Target Driven Instance Detector (TDID) presented in [53], an architecture designed to recognise and classify specific instances of object classes.”
BackgroundMethodInfluentialSpatial Commonsense Graph for Object Localisation in Partial Scenes
Giuliari, F., Skenderi, G., Cristani, M., Wang, Y., & Bue, A. D. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01891 · Cited by 27 · cited 2×
“Moreover, spatial relations among objects have also been formulated as graphs and embedded via Graph Convolutional Networks to guide the navigation policy [25], where external commonsense knowledge has also shown advantages for the object localisation via spatial graph learning [3].”
BackgroundSelf-citationGAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment
Ye, X., Lin, Z., Lee, J., Zhang, J., Zheng, S., & Yang, Y. (2019) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2019.2930426 · Cited by 24
Zero-shot Active Visual Search (ZAVIS): Intelligent Object Search for Robotic Assistants
Park, J., Yoon, T., Hong, J., Yu, Y., Pan, M. K. X. J., & Choi, S. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10161345 · Cited by 24 · cited 1×
“…2D grid map of the scene in our POMCP modelling: blue dots are the possible object locations, purple crosses are the possible robot poses. contexts, such as the domotic field [2], [3], [4], personal assistance [5], search and rescue [6], [7], and the very intriguing Mars exploration [8].”
BackgroundInfluence of State-Variable Constraints on Partially Observable Monte Carlo Planning
Castellini, A., Chalkiadakis, G., & Farinelli, A. (2019) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2019/769 · Cited by 22
Self-citationCastellini, A., Marchesini, E., & Farinelli, A. (2021) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2021.104382 · Cited by 22 · cited 1×
“Others[46],[47]concernrobotnavigationusingonly POMCPorhierarchicalmethodsapproacheswithPOMCPfor high-levelcontrolandneuralnetworksforlow-levelcontrol.”
Self-citationPOMP++: Pomcp-based Active Visual Search in unknown indoor environments
Giuliari, F., Castellini, A., Berra, R., Bue, A. D., Farinelli, A., Cristani, M., Setti, F., & Wang, Y. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9635866 · Cited by 18 · cited 2×
“To the best of our knowledge, the only works available in the literature about AVS with POMCP are [12], [45].”
BackgroundSelf-citationRisk-aware shielding of Partially Observable Monte Carlo Planning policies
Mazzi, G., Castellini, A., & Farinelli, A. (2023) · Artificial Intelligence · doi:10.1016/j.artint.2023.103987 · Cited by 17 · cited 1×
“Restrictions apply. approaches,[38],[39]allowstogenerateshieldsbasedonlogical rulestoimprovethesafetyofPOMCP.”
Self-citationPOMP: Pomcp-based Online Motion Planning for active visual search in indoor environments
Wang, Y., Giuliari, F., Berra, R., Castellini, A., Bue, A. D., Farinelli, A., Cristani, M., & Setti, F. (2020) · British Machine Vision Conference (BMVC) · doi:10.5244/c.34.190 · Cited by 15
Self-citationLearning Logic Specifications for Soft Policy Guidance in POMCP
Mazzi, G., Meli, D., Castellini, A., & Farinelli, A. (2023) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/janh8359 · Cited by 15
Self-citationLanguage-enhanced RNR-Map: Querying Renderable Neural Radiance Field maps with natural language
Taioli, F., Cunico, F., Girella, F., Bologna, R., Farinelli, A., & Cristani, M. (2023) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw60793.2023.00504 · Cited by 13 · cited 1×
“…2D grid map of the scene in our POMCP modelling: blue dots are the possible object locations, purple crosses are the possible robot poses. contexts, such as the domotic field [2], [3], [4], personal assistance [5], search and rescue [6], [7], and the very intriguing Mars exploration [8].”
BackgroundSelf-citationSchmid, J. F., Lauri, M., & Frintrop, S. (2019) · 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967805 · Cited by 12 · cited 8×
“We compare our proposed approach POMP-BE-PD against a random walk baseline –i.e., we allow the agent to randomly select an action among all the feasible ones at each time step–, and four state-of-the-art approaches, namely: EAT [10], DQN [54], DQN-TAM [55], and our previous work POMP [13].”
BackgroundMethodInfluentialLiu, S., Tian, G., Cui, Y., & Shao, X. (2022) · Frontiers of Information Technology & Electronic Engineering · doi:10.1631/fitee.2200109 · Cited by 12 · cited 1×
“We compare our proposed approach POMP-BE-PD against a random walk baseline –i.e., we allow the agent to randomly select an action among all the feasible ones at each time step–, and four state-of-the-art approaches, namely: EAT [10], DQN [54], DQN-TAM [55], and our previous work POMP [13].”
MethodSpatial Relation Graph and Graph Convolutional Network for Object Goal Navigation
Kiran, D. A. S., Anand, K., Kharyal, C., Kumar, G., Gireesh, N., Banerjee, S., Roychoudhury, R. D., Sridharan, M., Bhowmick, B., & Krishna, M. (2022) · IEEE International Conference on Automation Science and Engineering (CASE) · doi:10.1109/case49997.2022.9926534 · Cited by 12 · cited 1×
“Moreover, spatial relations among objects have also been formulated as graphs and embedded via Graph Convolutional Networks to guide the navigation policy [25], where external commonsense knowledge has also shown advantages for the object localisation via spatial graph learning [3].”
BackgroundRule-based Shielding for Partially Observable Monte-Carlo Planning
Mazzi, G., Castellini, A., & Farinelli, A. (2021) · Proceedings of the International Conference on Automated Planning and Scheduling · doi:10.1609/icaps.v31i1.15968 · Cited by 11 · cited 1×
“Restrictions apply. approaches,[38],[39]allowstogenerateshieldsbasedonlogical rulestoimprovethesafetyofPOMCP.”
Self-citationMazzi, G., Castellini, A., & Farinelli, A. (2021) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/zrpc8764 · Cited by 10
Self-citationLeveraging Commonsense for Object Localisation in Partial Scenes
Giuliari, F., Skenderi, G., Cristani, M., Bue, A. D., & Wang, Y. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3272523 · Cited by 6
Self-citationRobots Tackle DARPA Underground Challenge
Leslie, M. (2022) · Engineering · doi:10.1016/j.eng.2022.04.003 · Cited by 5 · cited 1×
“…2D grid map of the scene in our POMCP modelling: blue dots are the possible object locations, purple crosses are the possible robot poses. contexts, such as the domotic field [2], [3], [4], personal assistance [5], search and rescue [6], [7], and the very intriguing Mars exploration [8].”
BackgroundA High-efficient Training Strategy for Deep Q-learning Network Used in Robot Active Object Detection
Liu, S., & Tian, G. (2022) · IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems · doi:10.1109/cyber55403.2022.9907473 · Cited by 3 · cited 1×
“We compare our proposed approach POMP-BE-PD against a random walk baseline –i.e., we allow the agent to randomly select an action among all the feasible ones at each time step–, and four state-of-the-art approaches, namely: EAT [10], DQN [54], DQN-TAM [55], and our previous work POMP [13].”
MethodPouya, P., & Madni, A. (2021) · ASCEND 2021 · doi:10.2514/6.2021-4024 · Cited by 2 · cited 1×
“…2D grid map of the scene in our POMCP modelling: blue dots are the possible object locations, purple crosses are the possible robot poses. contexts, such as the domotic field [2], [3], [4], personal assistance [5], search and rescue [6], [7], and the very intriguing Mars exploration [8].”
BackgroundLearning State-Variable Relationships in POMCP: A Framework for Mobile Robots
Zuccotto, M., Piccinelli, M., Castellini, A., Marchesini, E., & Farinelli, A. (2022) · Frontiers in Robotics and AI · doi:10.3389/frobt.2022.819107 · Cited by 2 · cited 1×
“Others[46],[47]concernrobotnavigationusingonly POMCPorhierarchicalmethodsapproacheswithPOMCPfor high-levelcontrolandneuralnetworksforlow-levelcontrol.”
Self-citationLearning Logic Specifications for Soft Policy Guidance in POMCPPreprint
Mazzi, G., Meli, D., Castellini, A., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2303.09172 · Cited by 2
Self-citationScalable Safe Policy Improvement via Monte Carlo Tree Search
Castellini, A., Bianchi, F., Zorzi, E., et al. (2023) · Research Repository (Delft University of Technology) · Cited by 1
Self-citationZero-shot Active Visual Search (ZAVIS): Intelligent Object Search for Robotic AssistantsPreprint
Park, J., Yoon, T., Hong, J., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2209.08803 · Cited by 1
A Dataset for Developing and Benchmarking Active VisionPreprint
Ammirato, P., Poirson, P., Park, E., Košecká, J., & Berg, A. C. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1702.08272 · Cited by 1
Habitat-Matterport 3D Semantics DatasetPreprint
Yadav, K., Ramrakhya, R., Ramakrishnan, S. K., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2210.05633
Research works printed but not linked: 1
- Mazzi, Active generation of logical rules for POMCP shielding, 1696
Cited by
Language-enhanced RNR-Map: Querying Renderable Neural Radiance Field maps with natural language
Taioli, F., Cunico, F., Girella, F., Bologna, R., Farinelli, A., & Cristani, M. (2023) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw60793.2023.00504 · Cited by 13
“There is an extensive research area that focuses on how to build maps to aid navigation in indoor environments [1, 4, 19, 11, 15, 26].”
BackgroundSelf-citationTaioli, F., Zorzi, E., Franchi, G., Castellini, A., Farinelli, A., Cristani, M., & Wang, Y. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.01745 · Cited by 25
“Recent policies can be divided into two categories: training-based [10, 19, 31, 44, 52, 64] and zero-shot policies [11, 21, 54, 63, 65, 68, 71].”
Self-citationAlgorithmic Perspectives on Active Object Recognition Systems: A Survey
Wang, Y., Li, J., Du, Z., Xue, T., & Zhang, T. (2025) · Tsinghua Science & Technology · doi:10.26599/tst.2025.9010096
Self-citationAIDA: AI-Driven Intelligent Diagnostics and Analytics
Ruan, Z., Gobbo, S., Cima, L., Mondo, M., Sharifi, S., Munari, E., Scarpa, A., Khan, O. S., Ólafsson, S., Giachetti, A., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-11381-8_11
Self-citationWang, Y., & Tian, G. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.131538 · Cited by 1
Self-citation
Metadata
- Authors
- Taioli, F., Giuliari, F., Wang, Y., Berra, R., Castellini, A., Bue, A. D., Farinelli, A., Cristani, M., & Setti, F.
- Affiliations
- University of Verona (IT)
- Italian Institute of Technology (IT)
- Fondazione Bruno Kessler (IT)
- ORCID
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2024-12
- Volume
- 46
- Issue
- 12
- Pages
- 11047-11058
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- 1939-3539
- OpenAlex
- W4401990919
- Semantic Scholar
- ef7a7001c843104eb16b19575e1f9e822efa2d84
- Cited by
- 8
- References
- 60
- Citations by year
- 2026: 5 · 2025: 2
- FWCI
- 1.03
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion, cc-by)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- PORTO@Iris (submittedVersion, cc-by)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Robotics and Sensor-Based Localization
- Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
- Keywords
- object detection, unsupervised learning, partially observable Markov decision processes, belief updating, sensor faults, probabilistic detection, online motion planning, floor maps, RGB-D sensing
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.