PhD Oral Exam - Juqi Hu, Mechanical Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

The evolution of autonomous driving technologies is expected to bring positive changes to transportation systems, our cities and society. Despite significant advances in sensing, signal processing and control technologies, the computational power demand continues to be the primary stumbling factor for realizing autonomous vehicles (AVs) with full autonomy on the road. The development of such an autonomy in turn necessitates the syntheses of computationally efficient and reliable guidance, navigation and control (GNC) methodologies including efficient methods for state estimations, safe and comfortable trajectory planning, and robust trajectory tracking control.

This thesis aims to design and develop trajectory planning and tracking strategies integrating the available tire-road friction coefficient (TRFC) so as to enhance the safety and reliability of the AVs under varying road friction conditions. The TRFC estimation methods based on the lateral and longitudinal dynamic responses of the vehicle are initially developed. The estimated TRFC is subsequently integrated in the trajectory planning and tracking frameworks to achieve effective GNC of AVs under different road conditions. A two-stage hierarchical framework is firstly developed for estimating TRFC in a computationally efficient manner considering vehicle’s lateral dynamic responses to double (DLC) and single (SLC) lane-change maneuvers. The vehicle lateral velocity, and thereby the front- and rear-wheels’ side-slip angles, are estimated in the first stage using the extended Kalman filter (EKF), while tire forces and the TRFC are estimated in the second stage using the unscented Kalman filter (UKF).

Owing to the unreasonable interference to the vehicle motion caused by lateral directional maneuvers, an alternate two-stage TRFC estimation framework is developed on the basis of the longitudinal dynamics of the vehicle. A sequence of braking pressure pulses is designed in the first stage to identify desired minimal pulse pressure needed for reliable estimation of TRFC with minimal interference with the vehicle motion. This stage also provides a qualitative estimate of TRFC. In the second stage, a modified force observer based on the wheel rotational dynamics is developed for estimating the tire braking force. A constrained unscented Kalman filter (CUKF) algorithm is subsequently proposed to identify the TRFC for achieving rapid convergence and enhanced estimation accuracy.

A trajectory planning scheme integrating the estimated TRFC is subsequently developed for path-change maneuvers considering both the maneuver safety and the occupant’s perception of comfort. For this purpose, a 7th order polynomial function is constructed to ensure continuity of the planned trajectory up to the derivative of the acceleration (jerk). The friction-adaptive acceleration and speed-adaptive jerk limits are further defined and integrated in the framework to enhance occupant’s comfort and acceptance. It is shown that the planning algorithm reduces to identification problem to only the lane change duration given a constant forward speed. The effectiveness of the proposed planning strategy is examined for driving scenarios involving broad variations in vehicle speed and TRFC. The acceptable traceability of the planned lane change trajectories is further demonstrated through path tracking analysis of a full-vehicle model.

An adaptive model predictive control (MPC) tracking scheme is proposed for tracking the desired lane-change path considering wide variations in vehicle speed and TRFC. With integrated consideration of output weights in the cost function together with constraints on the magnitude of the outputs defined by the TRFC, the proposed MPC scheme required only lateral position for tracking the planned path. An adaptive trajectory tracking control scheme is also proposed for AVs operating at relatively low speeds. An interesting way of integrating adaptive control gains with consideration of steering saturation by using the backstepping technique is designed to enhance trajectory tracking while ensuring that the commanded inputs are within the input boundaries. It is shown that the proposed controller can make the closed-loop system approximately globally asymptotically stable even in the presence of steering saturation. The effectiveness of the proposed tracking control scheme is then verified experimentally under relative low speeds using a scaled model self-driving car (QCar) test platform. Both the simulation and experimental results show that the proposed control scheme yields accurate trajectory tracking without violating the input constraints.