CS 7638 - Artificial Intelligence for Robotics - Indiana Drones Project - Solved

In this project, you were chosen to part of a team of treasure hunters who are seeking to extract an invaluable piece of ancient treasure. Unfortunately for us, the treasure is located in a dense and dangerous jungle making a typical safari impossible. That’s where you come in. As a robotics expert, your mission should you choose to accept it, is to program a drone to map the jungle environment (using SLAM) and navigate it through the jungle to extract the treasure while avoiding tree crashes.

The drone has special sensors that are able to detect the trees relative locations and size. The sensors will provide the bearing in radians (i.e. angle) and distance to the trees’ center in meters (relative to the drone’s location and orientation) as well as the radius of the tree in meters. You will also be able to move the drone by directly specifying the next location (i.e. by specifying distance in meters and steering in radians relative to the drone’s previous location and orientation).

Part A (worth 60%) asks you to complete the SLAM class in the indiana_drones.py file. Your mission is to implement a SLAM module that calculates and reports your drone’s position after each measurement and movement (relative to the arbitrary (0,0) staring point). All you need to do is fill in the appropriate member functions. The testing suite will call those functions at appropriate times to process the measurements (using process_measurements) and movement (using process_movement) and grade your SLAM systems accuracy in estimating the drone’s location (using get_coordinates).

This part is only to test your SLAM module. You do not guide the drone in part A and do not extract any treasure. A series of pre-scripted movements are performed and those movements will be provided to you one by one. For each test case, you will recieve 50% credit if your SLAM system accurately estimates the position of your drone within a distance threshold and 50% of the credit if each of the landmark (tree) locations are within a distance threshold. Points are deducted for each innacuarcy.

Note :

1.    You will only see trees that are within your drone’s measurement sensor’s horizon (in meters). So previously unseen trees may appear in your measurements as you move through the environment and get closer to new trees.

2.    Both measurements and movements contain noise. Since this noise value is not provided, you will need to come up with a way to estimate the noise

3.    Your drone will never have access to the map of where the randomly scattered trees are located.

Part B (worth 40%) asks you to complete the IndianaDronesPlanner class in the indiana_drones.py file. Your commander will provide you with the location of the treasure. Your goal in this part of the project is to use your newly developed SLAM module in conjuction with a navigation algorithm (that you need to code) to navigate the drone to the treasure’s location while avoiding trees in its path and extracting the treasure when it is reached.

The output of your navigation algorithm will be one of two actions in the next_move function: namely extract and move. The extract action extracts the treasure at your location if it exists. The move action moves the drone by the distance and steering you prescribe.

When you issue your extract action you should supply 3 arguments total, including the treasure type (*) and current estimated location (x, y) of the drone as follows: extract * 1.5 -2.1 [command treasure_type x y]. The treasure will only be extracted if it is within the defined radius (0.25 meters), if not there will be a time penalty for extracting dirt.

You should specify the movement as follows: move 1 1.57. [command distance steering] which means the drone will turn counterclock-wise 90 degrees [1.57 radians] first and then move a distance of 1 meter.

For each test case, you will receive 50% of the credit by extracting the treasure within the time limit, and

50% of the credit by avoiding any crash with the trees. However, if the drone doesn’t extract the treasure within the time limit, there will be no credit given. For each crash with one tree (multiple crashes with the same tree will be counted as one crash for grading purpose), there will be a 25% credit deduction. For example, if the drone extracted the treasure within the time limit but crashed into one tree and one tree only, you will receive 75% of the credit.

Note :

1.    In this project we assume the drone is a point (even though in the visualization it occupies some area).

2.    There is penalty for each tree the drone crashes into (whenever the drone enters within the radius of the tree’s center/the canopy of a tree). Remember that the drone moves on a path, so even if the starting and ending points of your movement aren’t inside the tree’s radius, the path could intersect it, which would result in a penalty.

3.    Your drone has a maximum turning angle [in radians] and a maximum distance [in meters] that it can move each timestep [both passed using a parameter]. Movement commands that exceed these values will be ignored and cause the drone to not move.