Alonzo Kelly

Optimal Rough Terrain Trajectory Generation for Wheeled Mobile Robots

An algorithm is presented for wheeled mobile robot trajectory generation that achieves a high degree of generality and efficiency. The generality derives from numerical linearization and inversion of forward models of propulsion, suspension, and motion for any type of vehicle. Efficiency is achieved by using fast numerical optimization techniques and effective initial guesses for the vehicle controls parameters. This approach can accommodate such effects as rough terrain, vehicle dynamics, models of wheel-terrain interaction, and other effects of interest. It can accommodate boundary and internal constraints while optimizing an objective function that might, for example, involve such criteria as obstacle avoidance, cost, risk, time, or energy consumption in any combination. The algorithm is efficient enough to use in real time due to its use of nonlinear programming techniques that involve searching the space of parameterized vehicle controls. Applications of the presented methods are demonstrated for planetary rovers.

Toward Reliable Off Road Autonomous Vehicles Operating in Challenging Environments

The DARPA PerceptOR program has implemented a rigorous evaluative test program which fosters the development of field relevant outdoor mobile robots. Autonomous ground vehicles were deployed on diverse test courses throughout the USA and quantitatively evaluated on such factors as autonomy level, waypoint acquisition, failure rate, speed, and communications bandwidth. Our efforts over the three year program have produced new approaches in planning, perception, localization, and control which have been driven by the quest for reliable operation in challenging environments. This paper focuses on some of the most unique aspects of the systems developed by the CMU PerceptOR team, the lessons learned during the effort, and the most immediate challenges that remain to be addressed.

Reactive Nonholonomic Trajectory Generation via Parametric Optimal Control

There are many situations for which a feasible nonholonomic motion plan must be generated immediately based on real-time perceptual information. Parametric trajectory representations limit computation because they reduce the search space for solutions (at the cost of potentially introducing suboptimality). The use of any parametric trajectory model converts the optimal control formulation into an equivalent nonlinear programming problem. In this paper, curvature polynomials of arbitrary order are used as the assumed form of solution. Polynomials sacrifice little in terms of spanning the set of feasible controls while permitting an expression of the general solution to the system dynamics in terms of decoupled quadratures. These quadratures are then readily linearized to express the necessary conditions for optimality. Resulting trajectories are convenient to manipulate and execute in vehicle controllers and they can be computed with a straightforward numerical procedure in real time.

Obstacle detection for unmanned ground vehicles: a progress report

To detect obstacles during off-road autonomous navigation, unmanned ground vehicles (UGVs) must sense terrain geometry and composition (terrain type) under day, night, and low-visibility conditions. To sense terrain geometry, we have developed a real-time stereo vision system that uses a Datacube MV-200 and a 68040 CPU board to produce 256/spl times/240-pixel range images in about 0.6 seconds/frame. To sense terrain type, we used the same computing hardware with red and near infrared imagery to classify 256/spl times/240-pixel frames into vegetation and non-vegetation regions at a rate of five to ten frames/second. This paper reviews the rationale behind the choice of these sensors, describes their recent evolution and on-going development, and summarizes their use in demonstrations of autonomous UGV navigation over the past five years.

Rough Terrain Autonomous Mobility—Part 2: An Active Vision, Predictive Control Approach

Off-road autonomous navigation is one of the most difficult automation challenges from the point of view of constraints on mobility, speed of motion, lack of environmental structure, density of hazards, and typical lack of prior information. This paper describes an autonomous navigation software system for outdoor vehicles which includes perception, mapping, obstacle detection and avoidance, and goal seeking. It has been used on several vehicle testbeds including autonomous HMMWVs and planetary rover prototypes. To date, it has achieved speeds of 15 km/hr and excursions of 15 km.We introduce algorithms for optimal processing and computational stabilization of range imagery for terrain mapping purposes. We formulate the problem of trajectory generation as one of predictive control searching trajectories expressed in command space. We also formulate the problem of goal arbitration in local autonomous mobility as an optimal control problem. We emphasize the modeling of vehicles in state space form. The resulting high fidelity models stabilize coordinated control of a high speed vehicle for both obstacle avoidance and goal seeking purposes. An intermediate predictive control layer is introduced between the typical high-level strategic or artificial intelligence layer and the typical low-level servo control layer. This layer incorporates some deliberation, and some environmental mapping as do deliberative AI planners, yet it also emphasizes the real-time aspects of the problem as do minimalist reactive architectures.

Linearized Error Propagation in Odometry

The related fields of mobile robotics and ground vehicle localization lack a linearized theory of odometry error propagation. By contrast, the equivalent Schuler dynamics which apply to inertial guidance have been known and exploited for decades. In this paper, the general solution of linearized propagation dynamics of both systematic and random errors for vehicle odometry is developed and validated. The associated integral transforms are applied to the task of eliciting the major dynamic behaviors of errors for several forms of odometry. Interesting behaviors include path independence, response to symmetric inputs, zeros, extrema, monotonicity and conservation. Applications to systems theory, systems design, and calibration are illustrated.

A new approach to vision-aided inertial navigation

We combine a visual odometry system with an aided inertial navigation filter to produce a precise and robust navigation system that does not rely on external infrastructure. Incremental structure from motion with sparse bundle adjustment using a stereo camera provides real-time highly accurate pose estimates of the sensor which are combined with six degree-of-freedom inertial measurements in an Extended Kalman Filter. The filter is structured to neatly handle the incremental and local nature of the visual odometry measurements and to handle uncertainties in the system in a principled manner. We present accurate results from data acquired in rural and urban scenes on a tractor and a passenger car travelling distances of several kilometers.

Generating near minimal spanning control sets for constrained motion planning in discrete state spaces

We propose a principled method to create a search space for constrained motion planning, which efficiently encodes only feasible motion plans. The space of possible paths is encoded implicitly in the connections between states, but only feasible and only local connections are allowed. Furthermore, we propose a systematic method to generate a near-minimal set of spatially distinct motion alternatives. This set of motion primitives preserves the connectivity of the representation while eliminating redundancy - leading to a very efficient structure for motion planning at the chosen resolution.

Rough Terrain Autonomous Mobility—Part 1: A Theoretical Analysisof Requirements

A basic requirement of autonomous vehicles is that of guaranteeing the safety of the vehicle by avoiding hazardous situations. This paper analyses this requirement in general terms of real-time response, throughput, and the resolution and accuracy of sensors and computations. Several nondimensional expressions emerge which characterize requirements in canonical form.The automatic generation of dense geometric models for autonomously navigating vehicles is a computationally expensive process. Using first principles, it is possible to quantify the relationship between the raw throughput required of the perception system and the maximum safely achievable speed of the vehicle. We derive several useful expressions for the complexity of terrain mapping perception under various assumptions. All of them can be reduced to polynomials in the response distance.The significant time consumed by geometric perception degrades real-time response characteristics. Using our results, several strategies of active geometric perception arise which are practical for autonomous vehicles and increasingly important at higher speeds.

Toward Optimal Sampling in the Space of Paths

While spatial sampling of points has already received much attention, the motion planning problem can also be viewed as a process which samples the function space of paths. We define a search space to be a set of candidate paths and consider the problem of designing a search space which is most likely to produce a solution given a probabilistic representation of all possible environments. We introduce the concept of relative completeness which is the prior probability, before the environment is specified, of producing a solution path in a bounded amount of computation. We show how this probability is related to the mutual separation of the set of paths searched. The problem of producing an optimal set can be related to the maximum k-facility dispersion problem which is known to be NP-hard. We propose a greedy algorithm for producing a good set of paths and demonstrate that it produces results with both low dispersion and high prior probability of success.