Home Research Publications Resume Links

Since September 2010, I have been affiliated with the Department of Mechanical Engineering at the University of Delaware. My current research interests are in the area of dynamics and control with application to bio-inspired robotic systems, specifically monopedal, bipedal and quadrupedal robots. In addition to my work in legged locomotion, I have recently started to investigate problems pertaining to the dynamics of collective decision making in multi-agent systems. Prior to working with legged robots and decision making, I have worked on algorithms for path planning and obstacle avoidance for nonholonomic mobile manipulators.
Legged Locomotion
The main thrust of my recent research is the advancement of the state of the art of robotic legged locomotion. As an alternative to traditional wheeled and tracked ground vehicles, biologically-inspired legged systems are becoming increasingly common. There are many reasons for exploring the use of legs in locomotion. One reason is mobility. There is a need for vehicles that can travel over rough terrain, where wheeled vehicles cannot traverse. A second reason is to understand the underlying control mechanisms in human and animal locomotion. Concrete theories can guide biological research by suggesting models for experimental testing and verification, or can be used in biomedical engineering to improve the design of artificial limbs for the rehabilitation of leg amputees.
From a system-theoretic perspective, walking and running are equivalent to stabilizing distinguished periodic orbits corresponding to certain motion patterns of mechanical systems that are hybrid in nature, inherently underactuated, feature a multitude of constraints, and interact with highly unstructured, complex environments. While mathematical analysis has yielded important insight into the nature of legged systems, current control synthesis tools are still of limited use. This gap has led researchers to turn to more intuitive approaches, involving time-consuming trial-and-error techniques. Contrary to this trend, my ongoing research work intends to "bridge" this gap, by proposing a general framework for the systematic design of feedback control laws that work in concert with the natural compliant dynamics of the system to induce agile, elegant, provably stable motions in legged robots. This can be achieved by combining the practical advantages of compliant mechanisms with the analytical tractability offered by established nonlinear controller synthesis methods.

Representative publications
Thumper Scout II bounding
Decision Making
Choosing between two alternatives represents a large class of real-world decision-making problems faced by humans and animals in their natural environments. Two-Alternative Forced-Choice (TAFC) tasks offer the prospect of a principled understanding of the dynamics of such decision-making behaviors. This can be achieved through the introduction of mathematical models amenable to tractable analysis, which, under reasonable assumptions, can faithfully describe and predict key aspects of TAFC tasks.
Both behavioral and neural data provide evidence supporting the Drift-Diffusion Model (DDM) -a simple stochastic differential equation- as a plausible model for formally investigating the mechanisms governing simple TAFC tasks. In this work, we depart from the pure DDM representing a single decision-making unit, to consider the more general setting of multiple such units interconnected according to particular communication topologies. A collective decision making scenario is proposed and analyzed, according to which, multiple interconnected units, each represented by a DDM, accumulate evidence toward correctly identifying a (noisy) stimulus between two known alternatives. The objective is to investigate the effect of coupling among the decision-making units on the speed and accuracy of the decision. This work can be viewed as a first step toward a common framework for studying collective decision making in TAFC tasks in groups involving humans and engineered systems.
Forced response Undirected path

Representative publications
Nonholonomic Systems
In the past, I studied the kinematics, dynamics and control of robotic manipulators mounted on wheeled mobile vehicles. The behavior of these systems is governed by nonholonomic constraints, i.e. constraints containing velocities that cannot be integrated to yield constraints in positions, which significantly limit the paths that the system can follow e.g. a sideways motion of a car is not possible.
Our research efforts focused on the path planning and obstacle avoidance problem for wheeled mobile manipulators. Path planning is a central problem in robotics, and many elegant path planners have been proposed in the relevant literature. However, the majority of these planners do not apply to wheeled robots due to their nonholonomic nature. This work developed an obstacle avoidance methodology, which explicitly takes into account the nonholonomic constraints in a global sense, it does not require a priori knowledge of the feasible paths the system can follow, and it guarantees that the system will not encounter an obstacle during its motion. This is achieved through a mathematical transformation that brings the nonholonomic constraint to a form suitable for planning purposes. The proposed method is fast and efficient; it greatly reduces the amount of computations required for calculating an obstacle-free path, a fact that makes it very attractive, especially in real time applications, where on-board computational power is limited. The validity of the method has been tested by both computer simulations and physical laboratory experiments.
Mobile Manipulator Obstacle Avoidance

Representative publications