Ground vehicles with autonomous navigation require medium range external sensing for early obstacle detection and terrain mapping. Both are essential for path planning. The conventional method for accomplishing this is achieved by very complex and expensive LiDAR sensors with 32 to 64 individual lasers rotating rapidly and taking readings from the top of the vehicle. Our most pertinent research question however is to ascertain if a less cumbersome and cost effective setup of a small number of single beam laser rangefinder sensors can accomplish the same task. Our current method to achieve this is to sweep the sensors a variable number of steps along a predetermined angle. The information obtained is used to develop a complete and detailed view of the world around the autonomous vehicle. This paper is focused on a small laser ranger finder attached to a programmable pan-tilt mount positioned at the front of the autonomous vehicle. These were chosen as they are the most viable type of sensors for this process due to low interference and also because long range accuracy data can be obtained with reasonable power consumption. The sensors and mount are connected to a microcontroller which gathers position and distance information of significant objects in the path of the vehicle, in polar coordinates. This information is then converted to more useful Cartesian coordinates and plotted on a point cloud which is then used for path planning in real time. Future work will include sensor fusion of this system and an image sensor for detection of relevant objects such as traffic signs. Such systems establish a distance from the vehicle as well as distinguish them by shape and color.

No abstract

No abstract

Autonomous vehicles provide an opportunity to reduce highway congestion and emissions, while increasing highway safety. Intelligently routed vehicles will also be better integrated with existing traffic patterns, minimizing travel times. By reducing the time wasted in traffic; harmful emissions will consummately be reduced. Well-designed autonomous control systems provide for increased highway safety by reducing the frequency and severity of traffic accidents caused by driver error. In order to achieve this, a robust multi-layered control system must be designed, which minimizes the likelihood of computer error, while enabling seamless transition to and from human control. Autonomous vehicle navigation systems rely on accurate and timely sensor inputs to determine a vehicle’s location, attitude, speed, and acceleration. This paper describes a telemetry sensor fusion approach, which enables an autonomous vehicle to navigate, complex intersections, based on previously planned paths and near field sensors. This reduces computational overhead on the vehicle’s computer, and provides real time redundancy for system errors or delays. In conjunction with a full complement of environmental sensors, this path planning - path following approach enhances the robustness of autonomous vehicle operating models. This research supports the rapidly expanding field of autonomous automobiles by examining novel concepts for robust telemetry sensor fusion between inertial, GPS, and wheel speed sensors, which allows for error correction and enhanced positional accuracy, when compared to conventional navigation algorithms.