Christopher Vo

Behavior-based motion planning for group control

Despite the large body of work in both motion planning and multi-agent simulation, little work has focused on the problem of planning motion for groups of robots using external “controller” agents. We call this problem the group control problem. This problem is complex because it is highly underactuated, dynamic, and requires multi-agent cooperation. In this paper, we present a variety of new motion planning algorithms based on EST, RRT, and PRM methods for shepherds to guide flocks of robots through obstacle-filled environments. We show using simulation on several environments that under certain circumstances, motion planning can find paths that are too complicated for naïve “simulation only” approaches. However, inconsistent results indicate that this problem is still in need of additional study.

Scalable and robust shepherding via deformable shapes

In this paper, we present a new motion planning strategy for shepherding in environments containing obstacles. This instance of the group motion control problem is applicable to a wide variety of real life scenarios, such as animal herding simulation, civil crowd control training, and oil-spill cleanup simulation. However, the problem is challenging in terms of scalability and robustness because it is dynamic, highly underactuated, and involves multi-agent coordination. Our previous work showed that high-level probabilistic motion planning algorithms combined with simple shepherding behaviors can be beneficial in situations where low-level behaviors alone are insufficient. However, inconsistent results suggested a need for a method that performs well across a wider range of environments. In this paper, we present a new method, called DEFORM, in which shepherds view the flock as an abstracted deformable shape. We show that our method is more robust than our previous approach and that it scales more effectively to larger teams of shepherds and larger flocks. We also show DEFORM to be surprisingly robust despite increasing randomness in the motion of the flock.

DDDAMS-based Crowd Control via UAVs and UGVs

Unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) collaboratively play central roles in intelligence gathering and control in urban/border surveillance and crowd control. In this paper, we first propose a comprehensive planning and control framework based on dynamic-data-driven, adaptive multi-scale simulation (DDDAMS). We then discuss proposed algorithms enabling DDDAMS capability based on several methods such as 1) Bayesian-based information aggregation/disaggregation, 2) dynamic information updating based on observation/simulation, 3) temporal and spatial data fusion for enhanced performance, 4) multi-resolution strategy in temporal tracking frequency, and 5) cached intelligent observers. Finally, preliminary results based on the proposed framework, algorithms, and testbeds are discussed.

A DDDAMS-based UAV and UGV team formation approach for surveillance and crowd control

The goal of this paper is to study the team formation of multiple UAVs and UGVs for collaborative surveillance and crowd control under uncertain scenarios (e.g. crowd splitting). A comprehensive and coherent dynamic data driven adaptive multi-scale simulation (DDDAMS) framework is adopted, with the focus on simulation-based planning and control strategies related to the surveillance problem considered in this paper. To enable the team formation of multiple UAVs and UGVs, a two stage approach involving 1) crowd clustering and 2) UAV/UGV team assignment is proposed during the system operations by considering the geometry of the crowd clusters and solving a multi-objective optimization problem. For the experiment, an integrated testbed has been developed based on agent-based hardware-in-the-loop simulation involving seamless communications among simulated and real vehicles. Preliminary results indicate the effectiveness and efficiency of the proposed approach for the team formation of multiple UAVs and UGVs.

Cooperative coevolution and univariate estimation of distribution algorithms

In this paper, we discuss a curious relationship between Cooperative Coevolutionary Algorithms (CCEAs) and univariate Estimation of Distribution Algorithms (EDAs). Specifically, the distribution model for univariate EDAs is equivalent to the infinite population EGT model common in the analysis of CCEAs. This relationship may permit cross-pollination between these two disparate fields. As an example, we derive a new EDA based on a known CCEA from the literature, and provide some preliminary experimental analysis of the algorithm.

Following a Group of Targets in Large Environments

The problem of following multiple coherent targets using a camera is called the group following problem. While camera tracking is a popular subject in literature, the group following problem has not gained much attention despite that there are many scenarios where it is desired for a moving sensor to maximize its visibility of a group of moving targets. In this work, we address the scalability issue by investigating the idea of cached visibility. We will discuss two camera motion planners called cached intelligent observers (cio). In our experimental results, we show that cached visibility provides better balance between efficiency and performance than existing methods, particularly in large complex environments.

Following a large unpredictable group of targets among obstacles

Camera control is essential in both virtual and real-world environments. Our work focuses on an instance of camera control called target following, and offers an algorithm, based on the ideas of monotonic tracking regions and ghost targets, for following a large coherent group of targets with unknown trajectories, among known obstacles. In multiple-target following, the cameras primary objective is to follow and maximize visibility of multiple moving targets. For example, in video games, a third-person view camera may be controlled to follow a group of characters through complicated virtual environments. In robotics, a camera attached to robotic manipulators could also be controlled to observe live performers in a concert, monitor assembly of a mechanical system, or maintain task visibility during teleoperated surgical procedures. To the best of our knowledge, this work is the first attempting to address this particular instance of camera control.

Following multiple unpredictable coherent targets among obstacles

Camera control is essential in both virtual and real-world environments. The quality of the cameras placement and motion may spell the difference between usability and confusion. Our work focuses on an instance of camera control called target following, and offers an algorithm for following multiple targets with unpredictable trajectories, among known obstacles. To the best of our knowledge, this work is the first attempting to address this important problem.