Towards Intelligible Human-Robot Interaction: An Active Inference Approach to Occluded Pedestrian Scenarios

Kai Chen; Yuyao Huang; Guang Chen

cs.ROFeb 26, 2026

Towards Intelligible Human-Robot Interaction: An Active Inference Approach to Occluded Pedestrian Scenarios

Authors: Kai Chen, Yuyao Huang, Guang Chen

Organizations: Tongji University Shanghai, China

Abstract

The sudden appearance of occluded pedestrians presents a critical safety challenge in autonomous driving. Conventional rule-based or purely data-driven approaches struggle with the inherent high uncertainty of these long-tail scenarios. To tackle this challenge, we propose a novel framework grounded in Active Inference, which endows the agent with a human-like, belief-driven mechanism. Our framework leverages a Rao-Blackwellized Particle Filter (RBPF) to efficiently estimate the pedestrian's hybrid state. To emulate human-like cognitive processes under uncertainty, we introduce a Conditional Belief Reset mechanism and a Hypothesis Injection technique to explicitly model beliefs about the pedestrian's multiple latent intentions. Planning is achieved via a Cross-Entropy Method (CEM) enhanced Model Predictive Path Integral (MPPI) controller, which synergizes the efficient, iterative search of CEM with the inherent robustness of MPPI. Simulation experiments demonstrate that our approach significantly reduces the collision rate compared to reactive, rule-based, and reinforcement learning (RL) baselines, while also exhibiting explainable and human-like driving behavior that reflects the agent's internal belief state.

Figures & tables

Panel of Figure 1 — Figure 1. Reactive vs. Proactive Driving in a Critical Occluded Pedestrian Scenario. (a) A reactive agent ignores the uncertainty from occlusion, fails to slow down, and collides. (b) Our proactive agent reasons about the latent hazard, preemptively reducing speed to ensure a safe interaction with interpretable behaviors. Two diagrams comparing a reactive driving agent to a proactive one. The top diagram shows a reactive agent failing to slow down for an occluded pedestrian and colliding. The bottom diagram shows a proactive agent slowing down and swerving to safely avoid the same pedestrian.

Figure 2. Overview of the Active Inference Interaction Framework. Our framework enhances the core perception-action loop (top center) with two key mechanisms to facilitate a explainable interaction. For robust Belief Updating (top-right), Conditional Reset (bottom-left) prevents belief decay from occlusion. For proactive Planning (top-left), Hypothesis Injection (bottom-right) challenges the agent with worst-case scenarios, yielding cautious and belief-driven behavior. Overview of the Active Inference Interaction Framework.

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$0.905\pm 0.012$	$0.095\pm 0.012$	$\textbf{7.974}\pm\textbf{0.026}$	$3.679\pm 0.049$	$\textbf{1.739}\pm\textbf{0.041}$
0.1	0.8	$0.923\pm 0.011$	$0.077\pm 0.011$	$8.022\pm 0.029$	$3.857\pm 0.049$	$1.734\pm 0.036$
0.3	0.8	$0.947\pm 0.010$	$0.053\pm 0.010$	$8.653\pm 0.052$	$\textbf{3.890}\pm\textbf{0.041}$	$1.540\pm 0.030$
0.8	0.8	$\textbf{0.977}\pm\textbf{0.007}$	$\textbf{0.023}\pm\textbf{0.007}$	$9.148\pm 0.064$	$3.367\pm 0.024$	$0.963\pm 0.016$
0.3	0.0	$0.250\pm 0.018$	$0.750\pm 0.018$	$\textbf{6.448}\pm\textbf{0.007}$	$3.770\pm 0.084$	$1.220\pm 0.045$
0.3	0.2	$0.707\pm 0.019$	$0.293\pm 0.019$	$7.141\pm 0.029$	$3.572\pm 0.042$	$1.013\pm 0.025$

Table 1. Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Metrics are averaged across 5 scenarios).

Figure 3. Impact of Conditional Reset in an Occlusion Scenario. (a) Belief in the pedestrian’s existence, $B_{t}(z_{p})=\sum_{n}w_{t}^{(n)}\hat{z}_{p}^{(n)}$ , and (b) the agent’s speed over time. The agent without Conditional Reset (blue) allows its belief to decay to zero, leading to risky acceleration and a collision. In contrast, the agent with Conditional Reset (orange) maintains its belief, decelerates cautiously, and safely navigates past the revealed pedestrian to reach its goal.

Figure 4. Effect of Hypothesis Injection Ratio (

\rho_{H}

) on Behavior. (a) Agent’s trajectory and (b) speed profile for different

\rho_{H}

. A low ratio (

\rho_{H}=0.0

) leads to insufficient deceleration and a collision. As

\rho_{H}

increases, the agent plans a more cautious trajectory with a wider berth (a) by decelerating more significantly (b), thus successfully avoiding the pedestrian.

Figure 5. Impact of Initial Presence Belief (

B_{0}(z_{p})

) on Behavior. (a) Agent’s trajectory and (b) speed profile for different initial beliefs. A zero belief (

B_{0}(z_{p})=0.0

) leads to no evasive action and a collision. As the belief increases, the agent decelerates earlier and more significantly, ensuring a safe passage.

Panel of Table 2 — Table 2. Average Wall-clock Time (ms) per Planning Step for Varying Numbers of Particles ( $N$ ) and Candidate Policies ( $M$ ).

Figure 6. Qualitative Comparison of Driving Strategies. The trajectories show how each method responds to the pedestrian appearing at different times. Baselines execute fixed, non-adaptive strategies. In contrast, our method (bottom right) dynamically adapts its maneuver to the perceived risk, demonstrating intelligent behavior.

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.178\pm 0.016$	$0.822\pm 0.016$	$6.508\pm 0.042$	$3.050\pm 0.083$	$0.835\pm 0.071$
Rule-based	$0.587\pm 0.020$	$0.413\pm 0.020$	$7.755\pm 0.036$	$2.971\pm 0.043$	$0.650\pm 0.032$
PPO-LSTM	$0.725\pm 0.018$	$0.275\pm 0.018$	$\textbf{4.188}\pm\textbf{0.002}$	$\textbf{4.080}\pm\textbf{0.038}$	$\textbf{9.663}\pm\textbf{0.086}$
Ours	$\textbf{0.947}\pm\textbf{0.010}$	$\textbf{0.053}\pm\textbf{0.010}$	$8.653\pm 0.052$	$3.890\pm 0.041$	$1.540\pm 0.030$

Table 3. Quantitative Comparison with Baseline Methods (Metrics are averaged across 5 scenarios).

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$0.783\pm 0.038$	$0.217\pm 0.038$	$\textbf{7.685}\pm\textbf{0.056}$	$\textbf{3.806}\pm\textbf{0.125}$	$\textbf{1.798}\pm\textbf{0.113}$
0.1	0.8	$0.858\pm 0.032$	$0.142\pm 0.032$	$7.700\pm 0.061$	$3.762\pm 0.123$	$1.659\pm 0.105$
0.3	0.8	$\textbf{0.925}\pm\textbf{0.024}$	$\textbf{0.075}\pm\textbf{0.024}$	$8.172\pm 0.076$	$3.651\pm 0.093$	$1.519\pm 0.081$
0.8	0.8	$0.908\pm 0.026$	$0.092\pm 0.026$	$8.231\pm 0.101$	$3.659\pm 0.067$	$1.216\pm 0.043$
0.3	0.0	$0.433\pm 0.045$	$0.567\pm 0.045$	$\textbf{6.452}\pm\textbf{0.013}$	$\textbf{3.687}\pm\textbf{0.140}$	$1.173\pm 0.080$
0.3	0.2	$0.783\pm 0.038$	$0.217\pm 0.038$	$7.040\pm 0.057$	$3.537\pm 0.097$	$1.107\pm 0.056$

Appendix

Table C.1 . Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Turning-Back Pedestrian)

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$0.808\pm 0.036$	$0.192\pm 0.036$	$\textbf{7.721}\pm\textbf{0.052}$	$3.723\pm 0.072$	$1.799\pm 0.074$
0.1	0.8	$0.758\pm 0.039$	$0.242\pm 0.039$	$7.796\pm 0.058$	$4.144\pm 0.083$	$\textbf{1.838}\pm\textbf{0.066}$
0.3	0.8	$0.850\pm 0.033$	$0.150\pm 0.033$	$8.692\pm 0.125$	$\textbf{4.249}\pm\textbf{0.093}$	$1.637\pm 0.065$
0.8	0.8	$\textbf{0.983}\pm\textbf{0.014}$	$\textbf{0.017}\pm\textbf{0.012}$	$10.431\pm 0.139$	$3.278\pm 0.040$	$0.872\pm 0.031$
0.3	0.0	$0.167\pm 0.034$	$0.833\pm 0.034$	$\textbf{6.435}\pm\textbf{0.018}$	$3.141\pm 0.103$	$0.954\pm 0.153$
0.3	0.2	$0.558\pm 0.045$	$0.442\pm 0.045$	$7.057\pm 0.080$	$3.729\pm 0.071$	$1.100\pm 0.048$

Appendix

Table C.2 . Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Deceptive Accelerating Pedestrian)

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$0.933\pm 0.023$	$0.067\pm 0.023$	$8.248\pm 0.049$	$2.713\pm 0.030$	$0.939\pm 0.092$
0.1	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$\textbf{8.116}\pm\textbf{0.038}$	$2.961\pm 0.030$	$\textbf{1.212}\pm\textbf{0.066}$
0.3	0.8	$0.992\pm 0.008$	$0.008\pm 0.008$	$8.336\pm 0.051$	$\textbf{3.267}\pm\textbf{0.023}$	$1.056\pm 0.044$
0.8	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$8.363\pm 0.058$	$3.241\pm 0.017$	$0.828\pm 0.015$
0.3	0.0	$0.042\pm 0.018$	$0.958\pm 0.018$	$\textbf{6.420}\pm\textbf{0.033}$	$2.613\pm 0.143$	$0.490\pm 0.244$
0.3	0.2	$0.692\pm 0.042$	$0.308\pm 0.042$	$7.343\pm 0.062$	$3.000\pm 0.036$	$0.675\pm 0.035$

Appendix

Table C.3 . Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Sudden Stop Pedestrian)

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$\textbf{8.586}\pm\textbf{0.033}$	$3.213\pm 0.042$	$\textbf{1.636}\pm\textbf{0.063}$
0.1	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$8.790\pm 0.054$	$3.399\pm 0.042$	$1.511\pm 0.063$
0.3	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$9.832\pm 0.129$	$\textbf{3.661}\pm\textbf{0.051}$	$1.475\pm 0.058$
0.8	0.8	$0.992\pm 0.016$	$0.008\pm 0.008$	$9.445\pm 0.143$	$3.107\pm 0.022$	$0.730\pm 0.018$
0.3	0.0	$0.000\pm 0.000$	$1.000\pm 0.000$	$-$	$-$	$-$
0.3	0.2	$0.533\pm 0.046$	$0.467\pm 0.046$	$\textbf{7.461}\pm\textbf{0.103}$	$3.216\pm 0.058$	$0.571\pm 0.018$

Appendix

Table C.4 . Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Sudden Appearance Pedestrian)

$\rho_{H}$	$B_{0}(z_{p})$	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
0.0	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$\textbf{7.537}\pm\textbf{0.031}$	$4.913\pm 0.113$	$\textbf{2.496}\pm\textbf{0.033}$
0.1	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$7.610\pm 0.034$	$\textbf{5.076}\pm\textbf{0.111}$	$2.463\pm 0.033$
0.3	0.8	$0.967\pm 0.016$	$0.033\pm 0.016$	$8.186\pm 0.094$	$4.678\pm 0.107$	$2.039\pm 0.055$
0.8	0.8	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$9.227\pm 0.143$	$3.568\pm 0.078$	$1.183\pm 0.037$
0.3	0.0	$0.608\pm 0.045$	$0.392\pm 0.045$	$\textbf{6.451}\pm\textbf{0.010}$	$4.080\pm 0.121$	$1.377\pm 0.046$
0.3	0.2	$0.967\pm 0.016$	$0.033\pm 0.016$	$6.950\pm 0.029$	$4.114\pm 0.096$	$1.321\pm 0.048$

Appendix

Table C.5 . Influence of Initial Presence Belief (

B_{0}(z_{p})

) and Hypothesis Injection Ratio (

\rho_{H}

) (Hesitant Pedestrian)

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.325\pm 0.043$	$0.675\pm 0.043$	$6.441\pm 0.060$	$2.931\pm 0.098$	$0.932\pm 0.113$
Rule-based	$0.792\pm 0.037$	$0.208\pm 0.037$	$7.674\pm 0.074$	$3.327\pm 0.089$	$0.855\pm 0.058$
PPO-LSTM	$0.650\pm 0.044$	$0.350\pm 0.044$	$\textbf{4.172}\pm\textbf{0.005}$	$\textbf{3.772}\pm\textbf{0.087}$	$\textbf{8.493}\pm\textbf{0.400}$
Ours	$\textbf{0.925}\pm\textbf{0.024}$	$\textbf{0.075}\pm\textbf{0.024}$	$8.172\pm 0.076$	$3.651\pm 0.093$	$1.519\pm 0.081$

Appendix

Table D.1 . Quantitative Comparison with Baseline Methods (Turning-Back Pedestrian)

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.000\pm 0.000$	$1.000\pm 0.000$	$-$	$-$	$-$
Rule-based	$0.267\pm 0.040$	$0.733\pm 0.040$	$8.619\pm 0.130$	$2.557\pm 0.034$	$0.605\pm 0.073$
PPO-LSTM	$0.733\pm 0.040$	$0.267\pm 0.040$	$\textbf{4.178}\pm\textbf{0.005}$	$3.932\pm 0.058$	$\textbf{10.000}\pm\textbf{0.000}$
Ours	$\textbf{0.850}\pm\textbf{0.033}$	$\textbf{0.150}\pm\textbf{0.033}$	$8.692\pm 0.125$	$\textbf{4.249}\pm\textbf{0.093}$	$1.637\pm 0.065$

Appendix

Table D.2 . Quantitative Comparison with Baseline Methods (Deceptive Accelerating Pedestrian)

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.200\pm 0.037$	$0.800\pm 0.037$	$6.742\pm 0.027$	$2.361\pm 0.027$	$0.053\pm 0.007$
Rule-based	$0.892\pm 0.028$	$0.108\pm 0.028$	$7.496\pm 0.018$	$2.530\pm 0.007$	$0.260\pm 0.029$
PPO-LSTM	$0.833\pm 0.034$	$0.167\pm 0.034$	$\textbf{4.195}\pm\textbf{0.002}$	$\textbf{3.616}\pm\textbf{0.049}$	$\textbf{9.708}\pm\textbf{0.166}$
Ours	$\textbf{0.992}\pm\textbf{0.008}$	$\textbf{0.008}\pm\textbf{0.008}$	$8.336\pm 0.051$	$3.267\pm 0.023$	$1.056\pm 0.044$

Appendix

Table D.3 . Quantitative Comparison with Baseline Methods (Sudden Stop Pedestrian)

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.000\pm 0.000$	$1.000\pm 0.000$	$-$	$-$	$-$
Rule-based	$0.133\pm 0.031$	$0.867\pm 0.031$	$7.706\pm 0.209$	$2.771\pm 0.118$	$0.682\pm 0.136$
PPO-LSTM	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$\textbf{4.200}\pm\textbf{0.000}$	$\textbf{4.947}\pm\textbf{0.048}$	$\textbf{10.000}\pm\textbf{0.000}$
Ours	$\textbf{1.000}\pm\textbf{0.000}$	$\textbf{0.000}\pm\textbf{0.000}$	$9.832\pm 0.129$	$3.661\pm 0.051$	$1.475\pm 0.058$

Appendix

Table D.4 . Quantitative Comparison with Baseline Methods (Sudden Appearance Pedestrian)

Method	Pass Rate (PR) $\uparrow$	Collision Rate (CR) $\downarrow$	Pass Time (PT) $\downarrow$	Min. Distance (MD) $\uparrow$	Min. TTC (TTC) $\uparrow$
Reactive	$0.367\pm 0.044$	$0.633\pm 0.044$	$6.441\pm 0.081$	$3.531\pm 0.148$	$1.174\pm 0.094$
Rule-based	$0.850\pm 0.033$	$0.150\pm 0.033$	$7.839\pm 0.069$	$3.262\pm 0.103$	$0.877\pm 0.069$
PPO-LSTM	$0.408\pm 0.045$	$0.592\pm 0.045$	$\textbf{4.188}\pm\textbf{0.005}$	$3.659\pm 0.077$	$\textbf{10.000}\pm\textbf{0.000}$
Ours	$\textbf{0.967}\pm\textbf{0.016}$	$\textbf{0.033}\pm\textbf{0.016}$	$8.186\pm 0.094$	$\textbf{4.678}\pm\textbf{0.107}$	$2.039\pm 0.055$

Appendix

Table D.5 . Quantitative Comparison with Baseline Methods (Hesitant Pedestrian)

Autonomous Driving Planning

Autonomous driving planning selects safe, feasible vehicle actions and trajectories amid traffic, uncertainty, road rules, and changing goals.

See all 382 papers

Multimodal Belief-Space Covariance Steering with Active Probing and Influence for Interactive Driving
Feb 16, 2026Devodita Chakravarty, John Dolan, Yiwei LyuRisk-Sensitive Planning Autonomous Driving Planning
Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, India · Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, USA · Department of Computer Science and Engineering, Texas A&M University, College Station, TX, USA
Pedestrian-Aware LLM-Driven Behavioral Planning for Autonomous Vehicles
May 16, 2026Aidana Baimbetova, Haruki Yonekura, Hamada Rizk +1Pedestrian Intention Prediction Autonomous Driving Planning
1The University of Osaka, Japan · 2RIKEN Center for Computational Science, Japan · 3Tanta University, Egypt
Safe Driving in Occluded Environments
Oct 15, 2025Zhuoyuan Wang, Tongyao Jia, Pharuj Rajborirug +5Autonomous Driving
Department of Electrical and Computer Engineering, Carnegie Mellon Universty · Faculty of Medicine, King Mongkut’s Institute of Technology Ladkrabang, Thailand · Department of Mechanical Systems Engineering, Nagoya University, Japan

Feb 16, 2026J/K move · Enter open · S save

Multimodal Belief-Space Covariance Steering with Active Probing and Influence for Interactive Driving

Devodita Chakravarty, John Dolan, Yiwei Lyu

Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, India · Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, USA · Department of Computer Science and Engineering, Texas A&M University, College Station, TX, USA

Risk-Sensitive Planning Autonomous Driving Planning Decision-Making under Uncertainty Belief-Space Planning Active Perception

Open paper AskPDF arXiv

Autonomous driving in complex traffic requires reasoning under uncertainty. Common approaches rely on prediction-based planning or risk-aware control, but these are typically treated in isolation, limiting their ability to capture the coupled nature of action and inference in interactive settings. This gap becomes especially critical in uncertain scenarios, where simply reacting to predictions can lead to unsafe maneuvers or overly conservative behavior. Our central insight is that safe interaction requires not only estimating human behavior but also shaping it when ambiguity poses risks. To this end, we introduce a hierarchical belief model that structures human behavior across coarse discrete intents and fine motion modes, updated via Bayesian inference for interpretable multi-resolution reasoning. On top of this, we develop an active probing strategy that identifies when multimodal ambiguity in human predictions may compromise safety and plans disambiguating actions that both reveal intent and gently steer human decisions toward safer outcomes. Finally, a runtime risk-evaluation layer based on Conditional Value-at-Risk (CVaR) ensures that all probing actions remain within human risk tolerance during influence. Our simulations in lane-merging and unsignaled intersection scenarios demonstrate that our approach achieves higher success rates and shorter completion times compared to existing methods. These results highlight the benefit of coupling belief inference, probing, and risk monitoring, yielding a principled and interpretable framework for planning under uncertainty.

Abstract

Autonomous Driving Planning

Explore similar work

Multimodal Belief-Space Covariance Steering with Active Probing and Influence for Interactive Driving

Pedestrian-Aware LLM-Driven Behavioral Planning for Autonomous Vehicles

Safe Driving in Occluded Environments