Motion Planning Knowledge - Case Study Example

The physics of Robots from a Programmers’ View Introduction The world has seen proliferation in robot technology mirrored in real-world applications. Motion planning is an activity that entails the problem of computing physical actions to complete a specified task has spurred research in robotics realm. The theory in application of robots outpaces the practical implementation. The practical limitations are robots must deal with huge amounts of unclear sensor data, doubt, real-time demands, nonlinear dynamic effects, undefined models and unusual objective constraints and functions (Kris). Motion planning knowledge is commonly applied in vigorous robotics activities such as ladder climbing, development of intelligent user interface designs for human-operated robots; and the navigation in collision prone scenarios and events Motion planning Motion planning combines the state of the art acts of improving by expanding, enlarging and refining path planning and optimal control and to produce the required algorithmic baselines to tackle optimal control difficulties in some environments. Optimal Motion Planning involve generation of basic bounding capsules that implements distance constraints for an optimal control problem and thereby attain collision avoidance (Berns and Dillmann). Also it can be suggested that adoption of a two-stage framework for complex robots in optimal motion planning can provide a lasting solution. When applied this framework yields robot trajectories that are collision free. The generation of most dependable trajectories that conform to the environment constraints is ever present in robots world (industrial and humanoid). In robotics it is common to use a given specific technique for humanoid motion planning to allow dynamic tasks to be carried out. For our case, the technique employed encompasses a two-stage approach. The first stage involves kinematic and geometric planner that gives the trajectory for both parties (the humanoid body and the carried tangible and visible entity). Then in the second phase the dynamic perceptual structure generator outputs suitable dynamically unchanging walking motion that make it possible for the robot to ferry the object. The motion planner gives the trajectory on which the object is carried. If the goal state (desired trajectory) is not capable of being obtained, the planned motion is altered either completely or partly up to a time that dynamically feasible motion is found. Planning for such dynamic tasks is greatly emphasized as it has to deal with dissimilar physical parameters of the object and dynamic effect. This is overcome by employing a dynamic pattern generator together with a planner and some degree of approval for collision avoidance. Motion planning contribution and related work In the recent years Motion planning has become a major area in humanoid research. Andō et al. proposed a two-stage planning strategy in finding dynamic motion from the motion planner by using dynamic balance filter. This strategy is deployed in several key areas that include: • Multi-resolution DOF exploitation • Task-oriented motion planning • Full body posture control to attain collision-free navigation. This is normally achieved through separating the robot into either fixed, movable or free limb parts. • An integrated motion control strategy for running and walking • Mobility planning for humanoid robots in navigating through constrained areas with limited spaces by distinctively altering the humanoid locomotion modes of walking, crawling and sidestepping. • Footstep planning on irregular surface terrain and navigation in collision prone environments • Realistic animation and dynamic computation • Digital actor animation and integration of dynamics and motion planning In respect to motion planning Andō et al. proposes an integrated solution that addresses both the accomplishment of complex tasks such as concurrent dynamic manipulation and locomotion, And handling of dynamic, geometric and kinematic effects. More emphases is put on how to integrate dynamic simulation and task planning in the same framework, and to gain from the benefits resulting from path planning methods like diffusion and sampling. Motion planning for dynamic tasks The programmer has to address the algorithmic issues that cater for geometric and kinematic constraints (3D obstacle avoidance and object manipulation respectively). The aim is to get robust, dynamically executing paths that are collision free resulting in performance of simultaneous activities such as sidestepping and object manipulation. The two phase approach respectively plans for global path and path re-shaping. The created dynamic motion is examined so as to determine its accuracy, quality, or condition by the collision checker. In case of collision or possible collision scenario the motion planner alters the preceding motion plans. The change is enforced to the trajectory in accordance with the detected collision. Task planning. In task planning a motion plan is computed. This plan guarantees collision free movements for the object and the robot. The programmer models a geometric bounding box that represents the robot. Both the robot and the object are considered in computing configuration space that allow 9-dimension search states (6 and 3 dimensions for the object and box respectively). Next factor to consider is locomotion constraints that guarantee natural robot motions. There are steering methods that are used to create a roadmap to overcome these constraints. One such method involve the use of Dubins curves which are of straight line segments and arcs of a circle that are in chronological succession and accounts for natural forward motions. The programmer considers the robot direction as tangent to the path. The other method considers the robot as Omni-directional and computes a straight line separating the robot configurations. Code for roadmap is then created and enriched by using the Dubins curves. At this stage the programmer has created a 9- dimensioned space. Inverse kinematics are also employed for they compute complementary configurations of the humanoid arms. Path reshaping Upon checking output from the pattern generator a decision need to made to determine if the resulting robot motion is collision-free otherwise reshaping becomes inevitable if collision freeness is compromised. This strategy is successfully utilised in highly constrained spaces like in mechanical assembly model. Path reshaping focuses on keeping apart the parts of the motion plan that require improvement and adjustment suit the environment. Such situations match the configurations that are in close proximity to the obstacles. At this stage the programmer requires another kind of motion planning algorithm for controlling the clearance to obstacles. Diffusion probabilistic technique is used to continually reshape a path with potential collision into a collision free path. This method decomposes a path into a series of admissible sub paths that are connected in respect to the increasing and decreasing measures of clearance and penetration respectively. If the process does not give a collision free path, then the roadmap is modified and a new program is computed. More research effort need to be put in path search as there are no known methods that provide exhaustive representation. The path search requires the use of a dynamics filtering function throughout the path search phase. The Distance Metric for humanoid robot models puts more relative weights on the coordinates of links with higher mass and proximity. Further, and the computed trajectories have to be verified using a real robot hardware. Trajectory generation is achieved through smoothing process where raw paths are smoothened by creating several passes along the path length and replacing portions of the path among chosen pairs of configurations. Filtering on the other hand is used to give dynamically stable trajectory as the output. Pattern Generator for Dynamic Walking Johns and Taylor suggests a walking pattern generator that is founded on a preview control of zero moment point. Zero moment point allow generation of bipedal walking motion for discretional foot positioning. In specific the pattern generator takes foot placement and other parameters like double support ratio and footstep length. As a result the motion of the humanoid upper parts can be specified while walking since the pattern generator considers zero moment point based dynamically stable walking pattern. The pattern generator can also give walking patterns for climbing motions experienced in stairs. The inputs for the robot are line segments and arcs of a circle that enable accomplishment of Dubins curves. By utilising this feature the programmer is able to achieve walking and collision free motions. Obstacle and Dynamic Balance Constraints for Humanoid Robots. The complexity characterised in motion planning inhibit the use of complete algorithms that are restrained to small dimensional configuration spaces. As a result programmers have prompted to use heuristic algorithms that utilise randomization. Despite the fact that these methods are not complete, they are reliable as they are able to get paths in configuration spaces of extreme dimensions with high probability. In addition most modern algorithmic path planning methods and complex dynamic pattern generator available cover motion planning for dynamic tasks in robots. These are algorithms for computing dynamically-stable collision-free trajectories in order to achieve full body posture goals. This algorithm is universal as it can be utilised in any robot that is restrained to balance constraint particularly high-level control interfaces for automatically computing movements to solve complex undertakings for humanoid robots that encompass both obstacle collision avoidance and balance constraints. The limitations of the algorithm form the basis for further research as the implementation of the planner can only deal with few fixed positions. The capacity to produce strong effect of different configuration space distance metrics needs to be improved. On a final note more effort need to be put on improving the methods for integrating visual or tactile feedback. Conclusion Motion planning for humanoid robots navigating rough outdoor terrain and object manipulation is necessary for complete robot autonomy. It is also useful and of great importance in mechanism design simulation. However, a certain level of difficulty becomes evident as the humanoid mechanisms become complex. This is attributed by greater dimensional configuration space translating to complex geometric constraints and extended functional capabilities Work cited 1) Agullo, Miguel. LEGO Mindstorm Masterpieces. 1st ed. Rockland, Mass.: Syngress, 2003. Print. 2) Berns, Karsten, and R Dillmann. Proceedings Of The Fourth International Conference On Climbing And Walking Robots. 1st ed. Bury St Edmunds: Professional Engineering Pub., 2001. Print. 3) Andō, Noriaki et al. Simulation, Modeling, And Programming For Autonomous Robots. 1st ed. Berlin: Springer, 2010. Print. 4) Hamann, Heiko. Space-Time Continuous Models Of Swarm Robotic Systems. 1st ed. Dordrecht: Springer, 2010. Print. 5) Johns, Kyle, and Trevor Taylor. Professional Microsoft Robotics Developer Studio. 1st ed. Indianapolis, Ind.: Wiley Pub., 2008. Print. 6) Kris Hauser. Motion Planning for Real World Robots. CU Engineering Online, 18 Mar. 2014. Web. 5 May. 2014. ‹ http://vimeo.com/89520605/›. Read More