Online Monitoring for Safe Pedestrian-Vehicle Interactions
DOI: 10.1109/itsc45102.2020.9294366
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of ensuring safe interactions between autonomous vehicles and pedestrians in unstructured environments, such as pedestrian zones. The authors identify two primary obstacles to interactive autonomy: the high variability and unpredictability of human behavior, which complicates motion prediction, and the difficulty of providing formal safety guarantees for complex autonomous stacks. To mitigate accident risks while avoiding overly conservative performance, the study proposes an integrated framework that combines pedestrian intent estimation with a reachability-based online monitoring system. This approach aims to provide rigorous, near-real-time safety assessments for human-robot interactions. The methodology involves a complete autonomous stack implemented on a Polaris GEM electric vehicle equipped with LIDAR, radar, GPS, and cameras. Pedestrian intent is estimated using a multi-hypothesis particle filter combined with the Generalized Potential Field Approach (GPFA). This module predicts future trajectories by assigning particles to potential goal locations (e.g., crossing or walking parallel to the vehicle) and updating their weights based on noisy sensor measurements. For safety monitoring, the authors developed an Online Predictive Reachability Analysis (OPRA) module. Leveraging the DryVR tool, OPRA uses data-driven verification to compute over-approximations of reachable sets for the vehicle’s nonlinear dynamics. By partitioning initial state estimates and bloating simulation traces with a learned sensitivity function, the system achieves computation times of approximately 0.1 seconds for a 3–5 second look-ahead horizon. A decision module compares these vehicle reach sets against pedestrian trajectories; if an intersection with an unsafe set is detected, the controller switches from speed-tracking to braking mode. Experimental results demonstrate the feasibility and accuracy of the proposed system. In simulation and real-world tests, the intent estimation module accurately predicted pedestrian trajectories and goal probabilities. The OPRA module achieved high computational efficiency, with reach set computation times ranging from 0.096 to 0.163 seconds depending on the look-ahead duration. Accuracy evaluations showed that the computed reach sets contained the actual vehicle trajectories with 92% to 99% confidence, depending on the level of uncertainty modeled in the initial state estimates. The integrated system successfully navigated the test arena, maintaining safe separation from pedestrians by dynamically adjusting vehicle speed and steering based on the formal safety assessments. The significance of this work lies in its demonstration that formal verification methods, typically reserved for offline analysis, can be adapted for online, real-time safety monitoring in autonomous systems. By integrating data-driven intent prediction with reachability analysis, the authors provide a framework that offers formal safety guarantees without sacrificing the efficiency required for dynamic environments. This approach addresses the critical need for robust, verifiable safety mechanisms in autonomous vehicles operating alongside vulnerable road users, offering a pathway to reduce reliance on heuristic decision-making and improve overall system reliability.
Key finding
The integrated system successfully combines pedestrian intent estimation with online reachability analysis to provide formal safety assurances for autonomous vehicle-pedestrian interactions with near-real-time performance and high computational accuracy.
Methodology
simulation_modeling
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 7 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.