A Robotic Treasure Hunt Designed for UC Merced CSE 180. Progams written and designed by Adam Pluguez and Irian Aguiluz Due to the project being setup on each machine local text files in the build folder of the workspace need to be overwritten or have locations changed for the workspace to compile.
Problem description •Given the map of an indoor bounded environment, write a robot controller that finds an unknown number of tags hidden in the environment •tags are detected by a given sensor (logical camera — more later) •You must report (print to the screen): •how many tags you found •for each tag: identity (string), and absolute pose in the map.
Available nodes •amcl for localization •move_base to move the robot around •map_server to get the map of the environment •logical camera to detect objects in the environment
Map of the Environment •Can be read from disk or retrieved from map_server •robot produced. Mostly accurate but with some imprecision •your code must be generic with respect to the map, i.e., your code will be tested on a map different from the one you are given for testing •Pseudo info available either from disk (yaml file) or map_server
Logical Camera Description •Posts info on /objectsDetected •publishes messages of type message of type logical_camera_plugin/logicalImage •Example: modelName: couch_0 pose_pos_x: 1.47894906998 pose_pos_y: -1.03072357178 pose_pos_z: -0.368999242783 pose_rot_x: -1.08818733224e-06 pose_rot_y: -2.45950900535e-07 pose_rot_z: 0.705672979355 pose_rot_w: 0.708537697792
Strategy: Our strategy for finding the treasures was to create two separate nodes one for moving around the environment without colliding into obstacles and to other to continually be scanning for treasures.The movement node is called “gototreasure”, while the locator node is called ”treasurelocator”. The “gototreasure” node uses “move-base” to move to assigned goals. The dwa planner tolerances were adjusted so that the robot did not overcompensate trying to reach an exact point, as this was an obvious problem early on. The approach used to cover the entire map area involved assuming unknown obstacles randomly distributed. A efficient group of fixed points which would traverse the entire map with no obstacles was planned. These programmed points combined with the local planner should allow the robot to successfully traverse most if not all areas of a given map regardless of obstacle locations. The node “treasurelocator” is responsible for using the logical camera to scan for treasures, print out the treasure’s name, and coordinates in relation to the map. The logical camera gives the location of the treasure in relation to the robots current position in order to get the location of the treasure in relation to the map, first we use amcl get the location of the robot. From here we can use change of coordinates to get the estimated location of the treasure. Due to their being multiple treasures, in order to keep track of which treasures have already been found, a global vector was made containing the string id of each treasure. Since they all have a unique identifying name. Once it has been concluded that the treasure has not been found by checking the id in the vector, the new treasure’s id is placed in the vector, then its id and coordinates are printed. If the treasure has already been found before nothing is done and the robot keeps searching through the map for treasures. Grade given 85/100