Carnegie Mellon Robotics Institute
A challenging probem in Mechanical Engineering and robotics is the design of an outdoor mobile robot. The robot must successfully traverse through its environment, including rough terrain, grass, hills, etc, all while avoiding obstacles. For correct positioning of the robot, the wheels of the mobile robot must have good traction and never lose contact with the ground. For improved maneuverability (sometimes in tight spaces) the robot must also be able to rotate around a center point as close to its center of mass as possible. This also minimizes the use of energy when turning. All of these factors have been considered in the design of a second generation outdoor de-mining robot.
We decided on a rectangular frame for the robot. This allowed for easier positioning and accessibility of parts within the robot, which is useful in the development phase. The mobile mechanism is a differential-drive system: the two motors, controlled independently for forward, backward, and rotational movement, are located at the front of the robot. Having the wheels up front helps the robot drive onto a slope. A circular frame for the robot was considered and although it would allow the robot to rotate around its true center, such a frame, with the wheel axle running along the diameter of the frame would need support at the front and back of the robot. This system would have difficulty when moving from a flat terrain onto a slope or bumpy terrain. Since the wheels are placed at the front of the robot, we attached casters at the back for support.
Since the robot rotates around a center, frictional forces had to be considered. When deciding on a type of wheel to use, we chose pneumatic wheels that had a small width so as to reduce the frictional forces when the robot rotates. More importantly, friction on the free rotating casters can prohibit the robot from rotating properly and results in incorrect positioning. This occurs when the casters are positioned orthogonally to the direction of movement. We are considering ways of solving this problem. One solution was to actuate all four wheels and operate a skid-steering robot. This was not adequate because the skid-steering robot would experience lots of slippage when turning, thereby having an adverse effect on robot positioning with dead reckoning. Moreover, a skid-steering robot requires more power (again, to overcome friction when turning) and thus would have more expensive batteries and more weight. For these reasons, skid steering was ruled out. Another alternative was to actuate every drive wheel, as they do on the Nomad robot. This solution is incredibly maneuverable, both in rough terrains and tight spaces, but is expensive, weighty, and bulky because it requires many actuators. Our approach is to strike a balance among all of these constraints. Right now, we are considering an actuated caster. In this case a motor would control the positioning of the caster but would allow for a certain degree of free rotation. The second solution is a mechanical system that would react when an orthogonal force is applied causing the casters to rotate properly.
The robot is aware of its environment by using twenty-four sonar sensors. These sensors are positioned around the robot on a modular oval track. The sonars are equally spaced for more accurate readings. The front of the track forms a semicircle that scans 180 degrees and allows the robot to see in all directions. The module fits around the robot and is especially useful when replacing parts. We chose a modular approach to the sonar sensors so our sensing unit can be mounted on other robots.
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions