In this paper, we present the component technologies and the integration of these technologies for the development of an adaptive system of heterogeneous robots for urban surveillance. In our integrated experiment and demonstration, aerial robots generate maps that are used to design navigation controllers and plan missions for the team. A team of ground robots constructs a radio signal strength map that is used as an aid for planning missions. Multiple robots are to establish a mobile, ad-hoc communication network that is aware of the radio signal strength between nodes and adapts to changing conditions to maintain connectivity. Finally, the team of aerial and ground robots is able to monitor a small village, and search for and localize human targets by the color of the uniform, while ensuring that the information from the team is available to a remotely located human operator. The key component technologies and contributions include (a) mission specification and planning software; (b) decentralized control for navigation in an urban environment while maintaining communication; (c) programming abstractions and composition of controllers for multi-robot deployment; (d) cooperative control strategies for search, identification, and localization of targets; and (e) three-dimensional mapping in an urban setting.

MissionLab is a mission specification system that implements a hybrid deliberative and reactive control architecture for autonomous mobile robots. The user creates and executes the robot mission plans through its graphical user interface. As robot deployments become more common in highly stressful situations, such as in dealing with explosives or biohazards, the usability of their mission specification system becomes critical. To address this need, a mission-planning "wizard" has been recently integrated into MissionLab. By retrieving and adapting past successful mission plans stored in its database, this new feature is designed to simplify the user's planning process. The latest formal usability experiments, reported in this paper, testing for usability improvements in terms of speed of the mission planning process, accuracy of the produced mission plans, and ease of use is conducted. This paper introduces the mission-planning wizard, describes the usability experiments (including design), and discusses the results in detail.

As the capabilities, range of missions, and the size of robot teams increase, the ability for a human operator to account for all the factors in these complex scenarios can become exceedingly difficult. Our previous research has studied the use of case-based reasoning (CBR) tools to assist a user in the generation of multi-robot missions. These tools, however, typically assume that the robots available for the mission are of the same type (i.e., homogeneous). We loosen this assumption through the integration of contract-net protocol (CNP) based task allocation coupled with a CBR-based mission specification wizard. Two alternative designs are explored for combining casebased mission specification and CNP-based team allocation as well as the tradeoffs that result from the selection of one of these approaches over the other.

This paper explains an episodic-memory based approach for computing anticipatory robot behavior in a partially observable environment. Inspired by biological findings on the mammalian hippocampus, here, the episodic memories retain a sequence of experienced observation, behavior, and reward. Incorporating multiple machine learning methods, this approach attempts to help reducing the computational burden of the partially observable Markov decision process (POMDP). In particular, the proposed computational reduction techniques include: 1) abstraction of the state space via temporal difference learning; 2) abstraction of the action space by utilizing motor schemata; 3) narrowing down the state space in terms of the goals by employing instance-based learning; 4) elimination of the value-iteration by assuming a unidirectional-linear-chaining formation of the state space; 5) reduction of the state-estimate computation by exploiting the property of the Poisson distribution; and 6) trimming the history length by imposing the cap on the number of episodes that are computed. Furthermore, claims 5) and 6) were empirically verified, and it was confirmed that the state estimation can be in fact computed in an O(n) time (where n is the number of the states), more efficient than a conventional Kalman-filter based approach of O(n 2 ).

Georgia Tech, as part of DARPA's Tactical Mobile Robotics (TMR) Program, is developing a wide range of mission specification capabilities for the urban warfighter. These include the development of a range of easily configurable mission-specific robot behaviors suitable for various battlefield and special forces scenarios; communications planning and configuration capabilities for small teams of robots acting in a coordinated manner; interactive graphical visual programming environments for mission specification; and real-time analysis tools and methods for mission execution verification. This paper provides an overview of the approach being taken by the Georgia Tech/Honeywell team and presents a range of preliminary results for a variety of missions in both simulation and on actual robots.