Tasks like search‐and‐rescue and urban reconnaissance benefit from large numbers of robots working together, but high levels of autonomy are needed to reduce operator requirements to practical levels. Reducing the reliance of such systems on human operators presents a number of technical challenges, including automatic task allocation, global state and map estimation, robot perception, path planning, communications, and human‐robot interfaces. This paper describes our 14‐robot team, which won the MAGIC 2010 competition. It was designed to perform urban reconnaissance missions. In the paper, we describe a variety of autonomous systems that require minimal human effort to control a large number of autonomously exploring robots. Maintaining a consistent global map, which is essential for autonomous planning and for giving humans situational awareness, required the development of fast loop‐closing, map optimization, and communications algorithms. Key to our approach was a decoupled centralized planning architecture that allowed individual robots to execute tasks myopically, but whose behavior was coordinated centrally. We will describe technical contributions throughout our system that played a significant role in its performance. We will also present results from our system both from the competition and from subsequent quantitative evaluations, pointing out areas in which the system performed well and where interesting research problems remain. © 2012 Wiley Periodicals, Inc.

URBAN RECONNAISSANCE AND search-and-rescue missions are ideal candidates for multi-robot teams due to the potential hazard the missions pose to humans and the inherent parallelism that can be exploited by teams of cooperating robots. However, these domains also involve challenging problems due to having to work in complex, stochastic, and partially observable environments. In particular, non-uniform and cluttered terrain in unknown environments is a challenge for both state-estimation and control, resulting in complicated planning and perception problems. Limited and unreliable communications 62 COMMUNICATIONS OF THE ACM MARCH 2013 VOL. 56 ! NO. 3 key insights Human operators can help a robot team be more efficient and recover from errors. A good state estimate, in the form of a map, is the most critical piece of information for a team of robots-and the most difficult to obtain. Grand Challenge competitions like MAGiC highlight challenging open problems and provide a venue for evaluating new approaches.

This paper introduces the concept of a multi-robot community in which mul tiple robots must fulfill their individual tasks while operating in a shared environment. Unlike typical multi-robot systems in which global cost func tions are minimized while accomplishing a set of global tasks, the robots in this work have individual tasks to accomplish and individual cost functions to optimize (e.g. path length or number of objects to gather). A strategy is presented in which a robot may choose to aid in the comple tion of another robot's task. This type of "altruistic" action leads to evolving altruistic relationships between robots, and can ultimately result in a decrease in the individual cost functions of each robot. However, altruism with respect to another robot must be controlled such that it allows a relationship where both robots are altruistic, but protects an altruistic robot against a selfish robot that does not help others. A quantitative description of this altruism is presented, along with a law for controlling an individuals altruism. With a linear model of the altruism dy namics, altruistic relationships are proven to grow when robots are altruistic, but protect an altruistic robot from a selfish robot. Results of task planning simulations are presented that highlight the decrease in individual robot cost functions, as well as evolutionary trends of altruism between robots.

Abstract-Large position errors plague GNSS-based sensors (e.g., GPS) due to poor satellite configuration and multipath effects, resulting in frequent outliers. Due to quadratic cost functions when optimizing SLAM via nonlinear least square methods, a single such outlier can cause severe map distortions. Following in the footsteps of recent improvements in the robustness of SLAM optimization process, this work presents a framework for improving sensor noise characterizations by combining a machine learning approach with max-mixture error models. By using max-mixtures, the sensor's noise distribution can be modeled to a desired accuracy, with robustness to outliers. We apply the framework to the task of accurately modeling the uncertainties of consumer-grade GPS sensors. Our method estimates the observation covariances using only weighted feature vectors and a single max operator, learning parameters off-line for efficient on-line calculation.

Autonomous robots must be able to detect hazardous terrain even when sensor data is noisy and incomplete. In particular, negative obstacles such as cliffs or stairs often cannot be sensed directly; rather, their presence must be inferred. In this paper, we describe the height-length-density (HLD) terrain classifier that generalizes some prior methods and provides a unified mechanism for detecting both positive and negative obstacles. The classifier utilizes three novel features that inherently deal with partial observability. The structure of the classifier allows the system designer to encode the capabilities of the vehicle as well as a notion of risk, making our approach applicable to virtually any vehicle. We evaluate our method in an indoor/outdoor environment, which includes several perceptually difficult real-world cases, and show that our approach out-performs current methods.