Altruistic Autonomy: Beating Congestion on Shared Roads

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.

Emergent Prosociality in Multi-Agent Games Through Gifting

“…Opponent Modeling: Reasoning about opponent behavior can lead to more complex interactions. With opponent modeling, agents can estimate their opponents' policies, parameters, or updates in order to inform their own learning [Foerster et al, 2018a;Sadigh et al, 2018;Sadigh et al, 2016;Shih et al, 2021;Xie et al, 2020;Zhu et al, 2020]. While opponent modeling has shown promising results, it often provides approximation solutions that can be suboptimal.…”

“…With gifting, they gift each other early in training, which mitigates risk, and they learn to hunt together in a prosocial manner. leads to higher rewards only if agents cooperate, such as alleviating traffic congestion [Bıyık et al, 2018;Lazar et al, 2019], sustainably sharing limited resources [Hughes et al, 2018], and altruistic human-robot interaction [Shirado and Christakis, 2017]. In a game-theoretic framework, the class of Stag Hunt games is a well-known instance of the tradeoff between social cooperation and safety.…”

Incentivizing Efficient Equilibria in Traffic Networks with Mixed Autonomy

“…We can then write the latency as a function of vehicle flow as well as a binary argument s i , which indicates whether the road is congested [23], [30]: congested. An equilibrium has an associated equilibrium latency experienced by all selfish users.…”

“…As shown in Fig. 2, equilibria may have one road in free-flow and rest congested, or all may be congested [23], [30].…”

“…Remark 1. Note that, unlike in BNE, road m * EQ will not necessarily be in free-flow and m * EQ is not necessarily the minimum index such that all selfish traffic can be feasibly routed at Nash Equilibrium [30]. Further, different elements of the set BFNE can have different indices for longest equilibrium and longest used road.…”