Design and Implementation of a Heading Angle Controller for an Autonomous Ground Vehicle

 

With the advent of increasing attention to surveillance, search and rescue operations in hazardous environments, militant insurgency, mine detection, mine clearance, etc., the navigation of Autonomous Ground Vehicles (AGVs) has attracted much attention in recent years. Significant research and development programs have been established during the past few years to develop fully autonomous ground vehicles, but the technologies are not yet ready for immediate use. Continuum motion generation algorithms have considered that the feasible path for navigation of an AGV is a combination of either arcs of a circle or straight-line segments. In these methods, the steering angle was also assumed to change instantaneously during transition between consecutive path curvatures. Therefore, if the actual motion is admissible or even possible at all, a greater control effort is required at the execution level.

 

Also, in traditional method of path tracking, the steering angle was calculated based on the wheelbase of the vehicle and the path curvature or the look-ahead distance to minimize the orientation error. In these methods, the gain or look-ahead distance was independently tuned to be stable by conducting a wide range of experiments at various speeds. A kinematic bicycle model may be useful if there is no rapid change in path curvature to ignore the steering saturation and rate limits. However, the longitudinal speed of the vehicle is assumed constant to be for a kinematic bicycle model and additionally, it only considers the length of the vehicle. Particularly, no effort was made to model the mass, moment of inertia and wheel slippage.

 

Even if there are many software and mathematical models available in the literature to analyze the dynamic performance, it is always difficult to identify or collect the required vehicle parameters from the vehicle manufacturer for simulation. In analyzing the vehicle handling performance, a critical aspect is the use of an appropriate tire model that can accurately characterize the ground-wheel interaction and one associated challenge is to obtain the corresponding model parameters. Expensive test equipment is required to obtain the values of the tire model parameters including their cornering stiffness.

 

This thesis begins with addressing the inclusion of steering actuator dynamics to capture the transient response while tracking the desired heading angle. To navigate an Ackermann steered vehicle at lower speeds, a realistic Point-to-Point navigation algorithm was considered for an obstacle free path. A suitable control-oriented model that could accurately characterize the phenomenon of interest has been used to design the controller. A novelty in designing this control system was to include the actuator dynamics, as the response time to steer the front wheel is of the same order as that of the heading angle dynamics of the vehicle. By accounting the dynamic model of the vehicle, a significant source of error in heading angle would be eliminated at the planning level rather than reactively at the execution level. Also, this model included the tire cornering stiffness as an unknown parameter and two approaches were attempted to obtain its value.

 

Subsequently, a J-turn maneuver was carried out on a test vehicle equipped with an inertial measurement unit and a global positioning system to corroborate the mathematical model. The performance of two controllers (namely a classical transfer-function based controller and an optimal linear quadratic regulator) were evaluated using the IPG: CarMaker® simulation platform over a range of speeds. The transfer-function based controller was also implemented on the experimental test vehicle at low speeds (high-speed experimental implementation was not possible due to safety concerns). It was found that when the vehicle was moving at a longitudinal speed between 1.7 m/s to 3.8 m/s, a change of 20° in the heading angle was achieved within 3.6 s with 5 % steady-state error. The study was extended to perform a double lane change maneuver on the test vehicle. Furthermore, the effects of cornering stiffness and longitudinal speed were evaluated for better tracking performance. It was observed that the controller response was more sensitive to the cornering stiffness of the tire at higher speeds. It was found that the transfer-function based heading angle controller could provide a comparable performance with respect to the linear quadratic regulator, while keeping the sensing requirements to a minimum; thus, it was suitable for real time implementation on an AGV.