# We Need to Talk: Identifying and Overcoming Communication-Critical Scenarios for Self-Driving

arXiv: https://arxiv.org/abs/2305.04352 · HTML: https://mlsanity.com/paper/2305.04352/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 7 May 2023.

## Built on by (most cited first, 8 of 8)

- [STAMP: Scalable Task And Model-agnostic Collaborative Perception](https://mlsanity.com/paper/2501.18616/) (2025)
- [Towards Vehicle-to-everything Autonomous Driving: A Survey on Collaborative Perception](https://mlsanity.com/paper/2308.16714/) (2023)
- [LangCoop: Collaborative Driving with Language](https://mlsanity.com/paper/2504.13406/) (2025)
- [AirV2X: Unified Air-Ground Vehicle-to-Everything Collaboration](https://mlsanity.com/paper/2506.19283/) (2025)
- [V2X-PC: Vehicle-to-everything Collaborative Perception via Point Cluster](https://mlsanity.com/paper/2403.16635/) (2024)
- [Automated Vehicles Should be Connected with Natural Language](https://mlsanity.com/paper/2507.01059/) (2025)
- [SafeCoop: Unravelling Full Stack Safety in Agentic Collaborative Driving](https://mlsanity.com/paper/2510.18123/) (2025)
- [Multi-Agent Embodied Autonomous Driving (MAEAD): From V2X Information Exchange to Shared World Models](https://mlsanity.com/paper/2606.13840/) (2026)

## Abstract (arXiv)

In this work, we consider the task of collision-free trajectory planning for connected self-driving vehicles. We specifically consider communication-critical situations--situations where single-agent systems have blindspots that require multi-agent collaboration. To identify such situations, we propose a method which (1) simulates multi-agent perspectives from real self-driving datasets, (2) finds scenarios that are challenging for isolated agents, and (3) augments scenarios with adversarial obstructions. To overcome these challenges, we propose to extend costmap-based trajectory evaluation to a distributed multi-agent setting. We demonstrate that our bandwidth-efficient, uncertainty-aware method reduces collision rates by up to 62.5% compared to single agent baselines.