Robot navigation in dense crowds : statistical models and experimental studies of human robot cooperation

This thesis explores the problem of mobile robot navigation in dense human crowds. We begin by considering a fundamental impediment to classical motion planning algorithms called the freezing robot problem: once the environment surpasses a certain level of complexity, the planner decides that all forward paths are unsafe, and the robot freezes in place (or performs unnecessary maneuvers) to avoid collisions. Since a feasible path typically exists, this behavior is suboptimal. Existing approaches have focused on reducing predictive uncertainty by employing higher fidelity individual dynamics models or heuristically limiting the individual predictive covariance to prevent overcautious navigation. We demonstrate that both the individual prediction and the individual predictive uncertainty have little to do with this undesirable navigation behavior. Additionally, we provide evidence that dynamic agents are able to navigate in dense crowds by engaging in joint collision avoidance, cooperatively making room to create feasible trajectories. We accordingly develop interacting Gaussian processes, a prediction density that captures cooperative collision avoidance, and a "multiple goal" extension that models the goal driven nature of human decision making. Navigation naturally emerges as a statistic of this distribution.

Most importantly, we empirically validate our models in the Chandler dining hall at Caltech during peak hours, and in the process, carry out the first extensive quantitative study of robot navigation in dense human crowds (collecting data on 488 runs). The multiple goal interacting Gaussian processes algorithm performs comparably with human teleoperators in crowd densities nearing 1 person/m², while a state of the art noncooperative planner exhibits unsafe behavior more than 3 times as often as the multiple goal extension, and twice as often as the basic interacting Gaussian process approach. Furthermore, a reactive planner based on the widely used dynamic window approach proves insufficient for crowd densities above 0.55 people/m². We also show that our noncooperative planner or our reactive planner capture the salient characteristics of nearly any dynamic navigation algorithm. For inclusive validation purposes, we show that either our non-interacting planner or our reactive planner captures the salient characteristics of nearly any existing dynamic navigation algorithm. Based on these experimental results and theoretical observations, we conclude that a cooperation model is critical for safe and efficient robot navigation in dense human crowds.

Finally, we produce a large database of ground truth pedestrian crowd data. We make this ground truth database publicly available for further scientific study of crowd prediction models, learning from demonstration algorithms, and human robot interaction models in general.

Essays on cooperation and reciprocity

This dissertation comprises three essays that use theory-based experiments to gain understanding of how cooperation and efficiency is affected by certain variables and institutions in different types of strategic interactions prevalent in our society.

Chapter 2 analyzes indefinite horizon two-person dynamic favor exchange games with private information in the laboratory. Using a novel experimental design to implement a dynamic game with a stochastic jump signal process, this study provides insights into a relation where cooperation is without immediate reciprocity. The primary finding is that favor provision under these conditions is considerably less than under the most efficient equilibrium. Also, individuals do not engage in exact score-keeping of net favors, rather, the time since the last favor was provided affects decisions to stop or restart providing favors.

Evidence from experiments in Cournot duopolies is presented in Chapter 3 where players indulge in a form of pre-play communication, termed as revision phase, before playing the one-shot game. During this revision phase individuals announce their tentative quantities, which are publicly observed, and revisions are costless. The payoffs are determined only by the quantities selected at the end under real time revision, whereas in a Poisson revision game, opportunities to revise arrive according to a synchronous Poisson process and the tentative quantity corresponding to the last revision opportunity is implemented. Contrasting results emerge. While real time revision of quantities results in choices that are more competitive than the static Cournot-Nash, significantly lower quantities are implemented in the Poisson revision games. This shows that partial cooperation can be sustained even when individuals interact only once.

Chapter 4 investigates the effect of varying the message space in a public good game with pre-play communication where player endowments are private information. We find that neither binary communication nor a larger finite numerical message space results in any efficiency gain relative to the situation without any form of communication. Payoffs and public good provision are higher only when participants are provided with a discussion period through unrestricted text chat.