[Paper Review] Altruistic Autonomy: Beating Congestion on Shared Roads

Erdem Bıyık, Daniel A. Lazar+2 more|arXiv (Cornell University)|Oct 29, 2018

Transportation Planning and Optimization10 citations

TL;DR

This paper proposes a formal model of mixed-autonomy traffic flow using the fundamental diagram of traffic, introducing altruistic autonomy to reduce congestion. It develops polynomial-time algorithms for computing robust and altruism-optimized equilibria, showing that even minimal altruism can reduce average latency by a factor of four compared to worst-case selfish equilibria in simulations.

ABSTRACT

Traffic congestion has large economic and social costs. The introduction of autonomous vehicles can potentially reduce this congestion, both by increasing network throughput and by enabling a social planner to incentivize users of autonomous vehicles to take longer routes that can alleviate congestion on more direct roads. We formalize the effects of altruistic autonomy on roads shared between human drivers and autonomous vehicles. In this work, we develop a formal model of road congestion on shared roads based on the fundamental diagram of traffic. We consider a network of parallel roads and provide algorithms that compute optimal equilibria that are robust to additional unforeseen demand. We further plan for optimal routings when users have varying degrees of altruism. We find that even with arbitrarily small altruism, total latency can be unboundedly better than without altruism, and that the best selfish equilibrium can be unboundedly better than the worst selfish equilibrium. We validate our theoretical results through microscopic traffic simulations and show average latency decrease of a factor of 4 from worst-case selfish equilibrium to the optimal equilibrium when autonomous vehicles are altruistic.

Motivation & Objective

• To formalize the impact of autonomous vehicles on shared roads using the fundamental diagram of traffic.
• To design a robust Nash equilibrium algorithm that maintains performance under unforeseen demand increases.
• To model and optimize for varying degrees of altruism among autonomous vehicle users to minimize total network latency.
• To validate theoretical results through microscopic traffic simulations in realistic road networks.

Proposed method

• Develops a mixed-autonomy traffic flow model based on the fundamental diagram, linking vehicle density, flow, and latency.
• Introduces a robustness metric quantifying resilience to additional, unforeseen traffic demand in Nash equilibria.
• Proposes a polynomial-time optimization algorithm to compute the best robust Nash equilibrium (RBNE) minimizing total latency.
• Defines an altruism profile to represent varying willingness of autonomous vehicles to take longer routes.
• Designs a polynomial-time algorithm for optimal altruistic routing (BANE) that minimizes total experienced latency.
• Validates theoretical results using SUMO-based microscopic simulations with realistic vehicle dynamics and flow constraints.

Experimental results

Research questions

• RQ1 How does altruistic autonomy affect network-wide traffic latency in mixed-human/autonomous vehicle networks?
• RQ2 What is the theoretical improvement in latency when autonomous vehicles are allowed to be altruistic compared to purely selfish routing?
• RQ3 How can we compute a robust Nash equilibrium that maintains performance under unexpected increases in traffic demand?
• RQ4 What is the impact of varying altruism levels on overall network efficiency and congestion reduction?

Key findings

• Even with arbitrarily small altruism, total network latency can be unboundedly better than in the absence of altruism.
• The best selfish equilibrium can be unboundedly better than the worst selfish equilibrium, highlighting the importance of routing strategy.
• In simulations, optimal altruistic routing reduced average latency by a factor of four compared to the worst-case selfish equilibrium.
• The robust Nash equilibrium (RBNE) algorithm maximizes resilience to unforeseen demand while maintaining low latency, though simulation mismatches with theory occurred due to flow capacity discrepancies.
• Altruistic autonomous vehicles significantly reduce congestion, as shown by lower vehicle densities and higher average speeds in free-flow conditions.
• Increasing altruism levels further decreases overall latency, with simulations showing improved performance across multiple demand scenarios.