Mark E. Campbell

Autonomous driving in urban environments: approaches, lessons and challenges

The development of autonomous vehicles for urban driving has seen rapid progress in the past 30 years. This paper provides a summary of the current state of the art in autonomous driving in urban environments, based primarily on the experiences of the authors in the 2007 DARPA Urban Challenge (DUC). The paper briefly summarizes the approaches that different teams used in the DUC, with the goal of describing some of the challenges that the teams faced in driving in urban environments. The paper also highlights the long-term research challenges that must be overcome in order to enable autonomous driving and points to opportunities for new technologies to be applied in improving vehicle safety, exploiting intelligent road infrastructure and enabling robotic vehicles operating in human environments.

Cooperative Tracking Using Vision Measurements on SeaScan UAVs

A cooperative tracking approach for uninhabited aerial vehicles (UAVs) with camera-based sensors is developed and verified with flight data. The approach utilizes a square root sigma point information filter, which takes important properties for numerical accuracy (square root), tracking accuracy (sigma points), and fusion ability (information). Important augmentations to the filter are also developed for delayed data, by estimating the correlated processes, and moving targets, by using multiple models in a square root interacting multiple model formulation. The final form of the algorithm is general and scales well to any tracking problem with multiple, moving sensors. Flight data using the SeaScan UAV is used to verify the algorithms for stationary and moving targets. Cooperative tracking results are evaluated using multiple test flights, showing excellent results.

Multiple agent-based autonomy for satellite constellations

Multiple, highly autonomous, satellite systems are envisioned in the near future because they are capable of higher performance, lower cost, better fault tolerance, reconfigurability and upgradability. This paper presents an architecture and multi-agent design and simulation environment that will enable agent-based multi-satellite systems to fulfill their complex mission objectives, termed ObjectAgentTM. Its application is shown for a distributed aperture radar mission, although its applicability spans many types of missions. Required spacecraft functions, software agents, and multi-agent organisations are described for the radar mission, as well as their implementation. Agent-based simulations of mission case studies show the autonomous operation of the multi-agent architecture, which can then be used to build, evaluate and compare autonomous software architectures for multiple satellite systems.

Planning Algorithm for Multiple Satellite Clusters

A generalized planning methodology for satellite clusters is proposed. The methodology utilizes Hamilton ‐ Jacobi‐Bellman optimality (minimum time or minimum fuel ) to generate quickly a set of maneuvers from an initial stable formation to a e nal stable formation. Maneuvers are selected from the original set based on the maneuver time, fuel, and collision proximity. The e nal maneuvers are calculated by optimizing the switch times using a realisticset of orbital dynamics. The algorithm is developed to be distributed and scaleswell as the number of satellites increases. A minimal level of communication is used because only switch times and collision proximity information are distributed from the planner. An example with four satellites maneuvering in an eccentric orbit (e=0.2) is presented. Results show that optimal cluster maneuvers (minimum time or minimum fuel ) can be generated within minutes, and most of the computational implementation can be accomplished in parallel. I. Introduction S ATELLITE clusters are envisioned as an enabling technology for defense- and science-based missions. NASA’ s Origins program is planning a series of missions that perform spaceborne interferometry to image far off planets for possible life forms. 1 The U.S. Air Force is planning a distributed space-based, synthetic aperture radar mission within the next few years, possibly followed by a full deployment. 2 In each case, clusters of satellites hold the promise of increasing performance and reliability through distribution, while decreasing cost. The latter is a key aspect that will rely on levels of autonomous control algorithms and software currently being developed.

Optimal Cooperative Reconnaissance Using Multiple Vehicles

A study on multivehicle trajectory planning for cooperative reconnaissance problems is presented. Specifically, this work develops understanding and insights into how vehicles cooperate in reconnaissance type missions in which target information is maximized. The performance metric used to guide the cooperation study is the amount of information, defined using the Fisher information matrix, that the sensing vehicles gather over their planned trajectory. A receding horizon optimal control formulation is developed and solved for trajectories that yield maximum information. High-risk zones and vehicle/terminal constraints are also are considered. Trends include the following: 1) vehicles with nonsymmetric sensors tend to triangulate as they get close to the target; 2) vehicles tend to move toward stationary targets as quickly as possible; 3) the addition of a third vehicle exhibits at least 50% less performance improvement than the addition of the second vehicle, and even less for nonsymmetric sensors; 4) optimization for multiple vehicles and targets is a strong function of target to target distances and sensor uncertainty symmetry; and 5) short planning horizons are preferable for moving targets.

Six-axis vibration isolation system using soft actuators and multiple sensors

Several types of Stewart platforms have been implemented by research groups to examine design and control issues in six-axis vibration isolation for space-based systems. Hood Technology Corporation and the University of Washington have taken the lessons learned from these various designs and developed a new hexapod that addresses the requirements of the Jet Propulsion Laboratorys planned spaceborne interferometry missions. This system is unique in its very soft axial stiffness (3-Hz corner frequency) for active isolation and pointing control, custom-designed voice coil actuator with a large displacement capability, and elastomeric flexures both for guiding the actuator and providing pivots at the end of each strut. In addition, there are four sensors in each strut for control topology design and evaluation. An overview of this unique six-axis isolator design and a summary of the control results for various sensor topologies, including multisensor and frequency-weighted isolation and pointing control, are presented. Controllers that experimentally achieved 20-25-dB reduction in vibration in all six degrees of freedom across the bandwidth of interest (5-20 Hz) are shown.

A Vision Based Geolocation Tracking System for UAV ’ s

The design and implementation of a vision based geolocation tracking system for uninhabited aerial vehicles (UAV’s) is described. The UAV components include the avionics, ground station, and a gimballing camera with feedback isolation and command loops. The point of interest is locked in the camera image by maintaining the center of the image frame-to-frame. An architecture for a geolocation tracking estimator is developed and demonstrated. The estimator has the unique characteristics of being 1) modular so that it can work with different camera systems and different UAV components; 2) able to compensate for random and bias uncertainties in the UAV hardware components; 3) able to run in real time on a relatively slow processor; and 4) able to deliver an accurate estimate of the location and uncertainty of the object being tracked. Flight results using the SeaScan UAV show excellent results for two and three dimensional tracking of stationary and moving targets.

Optimal Planner for Spacecraft Formations in Elliptical Orbits

An optimal planner for spacecraft formations in elliptical reference orbits is presented. A fast solution to an individual spacecraft minimum time or fuel maneuver using the Hamilton-Jacobi-Bellman formulation is developed using spline approximations to evaluate thrust effect integrals. The individual optimal spacecraft maneuvers use realistic low-thrust, bounded inputs similar to future electric propulsion systems. A formation optimal planner is then formulated using the individual spacecraft maneuvers as a basis. The formulation is easily scalable to larger clusters, provably optimal over the formation, and numerically robust; it also requires minimal communication between fleet members. An example is presented of a tetrahedron formation in a highly elliptical reference orbit (e = 0.8), with solutions to both formation minimum-time and minimum-fuel problems given. Comparison with linear programming techniques show a distinct savings in fuel usage for high-eccentricity examples.

Estimation architecture for future autonomous vehicles

An architecture for the development of online models to support future uninhabited aerial vehicles is developed. The architecture is based on a new filter, called the unscented Kalman filter, that approximates the state and noise stochastic distributions, rather than the dynamics. A square root version of the unscented Kalman filter is shown to have better characteristics for online implementation than traditional methods, such as less sensitivity to tuning, initial conditions, and sample frequency. The estimation methodology is shown to be able to estimate the nonlinear state and model parameters for an aircraft during failure, and to generate aerodynamic models with potential application to online control reconfiguration.

Online nonlinear guaranteed estimation with application to a high performance aircraft

The most commonly used filters (or estimators) for online control customization are based upon stochastic assumptions on the noise. In this paper a set-membership filter is extended to be used with nonlinear systems for which the Jacobian and Hessian are continuous over the uncertainty interval. This filter relies on assuming that the noise sources are bounded in order to obtain hard bounds on state and parameter estimates. These bounds are then compatible with robust control design so that the controller can be adequately updated in real-time. The method proposed is referred to as an extended set-membership filter and is applied to a two-state example and to the state and parameter estimations of a complex F-15 like model.