Active Sensing And Estimation For Nonlinear Systems With Applications To Car-Bicycle Collision Prevention

Abstract

This dissertation develops a smart bicycle with novel sensors, active sensing control systems and nonlinear estimation algorithms for prevention of car-bicycle collisions. The smart bicycle tracks nearby cars, predicts their trajectories, and in the case of an anticipated collision with the bicycle sounds an audio alarm to alert the motorist to the presence of the bicycle. The challenges in developing the proposed system come from the inability to utilize the expensive radar and high-density LIDAR sensors typically used on prototype autonomous cars. Alternate sensor systems that are more than one order of magnitude cheaper must be used for the bicycle application. Active control of real-time sensor orientation and sophisticated estimation algorithms based on hybrid nonlinear observers are needed to enable the bicycle to reliably track nearby vehicles using the new inexpensive sensors. While the new active sensing controllers and nonlinear estimation algorithms were primarily developed and demonstrated on a bicycle in this project, they also have a large number of future applications on other vehicle platforms. First, the dissertation develops an active sensing system for a bicycle to accurately detect and track vehicles behind the bicycle in order to prevent rear-end collisions. A single beam laser sensor that is inexpensive, small and lightweight is adapted for this sensing mission. The rotational orientation of the laser sensor needs to be actively controlled in real-time in order to continue to focus on a rear vehicle, as the vehicle’s lateral and longitudinal distances from the bicycle change. This tracking problem requires controlling the real-time angular position of the laser sensor without knowing the future trajectory of the vehicle. The challenge is addressed by mounting the sensor on a rotationally controlled platform, using a novel receding horizon framework for active sensor orientation control, and an interacting multiple model framework for estimation. The features and benefits of this active sensing system are analyzed using simulation results. Then, extensive experimental results are presented using a prototype instrumented bicycle to show the performance of the system in detecting and tracking rear vehicles during both straight and turning maneuvers. Next, the topic of simultaneously searching for and tracking of multiple vehicles behind the bicycle using the same single beam laser sensor is addressed. Vehicles in the bicycle’s lane and in the adjacent left lane are both considered. The tasks involved are searching both lanes to detect presence of vehicles, tracking a vehicle’s trajectory once it has been detected, and switching between searching and tracking as needed. A rigorous search algorithm that minimizes the number of sensor rotational angles needed to search the entire region of interest is developed. An error covariance matrix approach is utilized to switch between tracking vehicles and searching a region of interest. Detailed simulation results are presented to show how the developed system handles the absence and presence of vehicles in the two lanes and handles different types of lane change maneuvers while tracking multiple vehicles. For tracking vehicles in front of the bicycle, including vehicles at traffic intersections, low-density LIDAR or radar sensors with a capability to make multiple distance measurements need to be utilized. The vehicle tracking task is challenging, especially due to the complex maneuvers and large orientation changes that occur with traffic at an intersection. A nonlinear observer is developed and utilized for the vehicle tracking task. Previous results in the literature on vehicle tracking have typically used an interacting multiple model estimator that needs different models for different modes of vehicle motion. This work uses a single nonlinear vehicle model that can be used for all modes of vehicle motion. Previous nonlinear observer design results from literature do not work for the nonlinear system under consideration due to the wide range of operating conditions that need to be accommodated. Hence, a new hybrid nonlinear observer design technique that utilizes switched gains and better bounds on the coupled nonlinear functions in the dynamics is developed. The asymptotic stability of the observer is established using Lyapunov techniques. The observer design with the developed technique is then implemented in both simulations and experiments. Experimental results show that the observer can simultaneously and accurately estimate longitudinal position, lateral position, velocity and orientation variables for the vehicle from radar measurements. Results are presented both for tracking of vehicle maneuvers on highways and of maneuvers on urban roads at traffic intersections where turns and significant changes in vehicle orientation can occur. The hybrid nonlinear observer developed for vehicle tracking also leads to a new conclusion on the importance of switched gains for observers in the case of non-monotonic nonlinear systems. A theoretical result which shows the non-existence of a stabilizing observer gain for non-monotonic systems, no matter how small the Lipschitz constant or Jacobian bounds of the nonlinearity, is developed. This is followed by development of a theoretical observer design technique based on switched observer gains for observer stabilization in non-monotonic nonlinear systems. Finally, a robust prototype bicycle with hardware for tracking of rear and side vehicles is developed in order to conduct a field operational test (FOT) in collaboration with a Minnesota bicycle company called Quality Bicycle Products (QBP). Ten employees of QBP will ride 10 instrumented bicycles for 6 months in two stages of the FOT. Data collected from the FOT will help refine the developed smart bicycle and evaluate its effectiveness in real world riding. The contributions of this dissertation include development of a smart vehicle tracking system suitable for on-bicycle implementation based on a new paradigm of active sensing and hybrid nonlinear estimation algorithms, development of a fundamental new observer design method for nonlinear non-monotonic systems and potential future commercialization of a new bicycle safety technology based on the outcomes of a proposed field operational test.