Before we could see “crash reported ahead” or “Police car ahead” on map apps, truckers had their own real-time information-sharing network. I assume they still use their radios to connect with other truckers on the road in real-time.

That kind of real-time information sharing should (will?) become standard in approximately ten years. The difference is, it will all be automated, with onboard computers doing most of the sharing, unless a human intervention is required.

Essentially, each car can be a moving computing unit, connected with others.

Tesla’s plan to implement a feature where every Tesla can “talk” to the other one is stage 1. At some point, interoperability will come into play as well. The evolution will be a multi-step process.

Here is an interesting research in this domain, though a bit old and focused on self-driving vehicles. https://lnkd.in/gC-rq6Vx

But this is a good example of a feature we would want with that network of interconnected vehicles. Irrespective of whether the vehicle is fully autonomous or has co-pilot capabilities.

As you can imagine, even with cars without self-driving, such a feature can help prevent a pile-up. A car involved in a crash immediately starts pinging any car approaching it, within a certain radius.

But let us explore this research for now.

The paper surveys strategies and research on collision avoidance in automated vehicles (AVs), especially in multiple-vehicle scenarios (chain collisions, rear-end crashes, etc.), not just single-vehicle avoidance.

It develops a taxonomy of existing methods/techniques, examines challenges, and identifies gaps. Then it proposes an AI‐enabled conceptual framework for dealing with “Multiple Vehicle Cooperation and Collision Avoidance” (MVCCA) in AVs.

To address gaps, the authors propose a framework with five phases:

Perception Phase: Advanced detection of surrounding vehicles, mapping, object refinement, risk indexing (recognizing whether only one vehicle nearby or several) etc.

Communication & Coordination Phase: Leveraging V2V / V2X / 5G or other low latency links to share information, allowing cooperative maneuvers like platooning, merging, etc.

Threat Assessment Phase: Classifying risk levels (regular, higher, critical) and estimating when chain collisions are likely so the system can proactively act.

Decision-Making Phase: Using learning-based (e.g. reinforcement learning, multi-agent RL) or hybrid methods to decide maneuvers that avoid or mitigate collisions.

Vehicle Control Phase: Executing braking, steering, and trajectory control in real time, under constraints (kinematic, dynamic) and accounting for uncertainties.