Requirements & Challenges

Requirements for a Passing Project Submission

The inset image at the bottom right when you're running in autonomous mode is packed with information. In this image, your map of navigable terrain, obstacles and rock sample locations is overplotted with the ground truth map. In addition, some overall statistics are presented including total time, percent of the world you have mapped, the fidelity (accuracy) of your map, and the number of rocks you have located (mapped) and how many you have collected.

The requirement for a passing submission is to map at least 40% of the environment at 60% fidelity and locate at least one of the rock samples (note: you're not required to collect any rocks, just map the location of at least 1). Each time you launch the simulator in autonomous mode there will be 6 rock samples scattered randomly about the environment and your rover will start at random orientation in the middle of the map.

The project is setup with lots of scaffolding code and you should only have to make minor modifications to the perception_step() and decision_step() functions (located within perception.py and decision.py, respectively) in order to arrive at a passing submission. If you're up for a challenge, however, you can take your project to a whole other level!

If you're struggling to get started on this project, or just want some help getting your code up to the minimum standards for a passing submission, we've recorded a walkthrough of the basic implementation for you but spoiler alert: this Project Walkthrough Video contains a basic solution to the project!.

The Udacity Sample Return Robot Challenge

The goal in the NASA Sample Return Challenge was not just to locate samples of interest but also to pick them up and return them to the starting point! In this Udacity challenge your goal is to map the entire environment, recover all the rock samples and return them to the starting point.

1. Optimizing Metrics:
The primary metrics of interest (shown in the image above) are time, percentage mapped, fidelity and number of rocks found. In the best case scenario, you would map the entire environment at a very high fidelity and locate and collect all six rock samples in the minimum total amount of time. To do this you'll need to not only optimize the accuracy your mapping analysis, but also the efficiency with which you traverse the environment.

Optimizing Map Fidelity Tip: Your perspective transform is technically only valid when roll and pitch angles are near zero. If you're slamming on the brakes or turning hard they can depart significantly from zero, and your transformed image will no longer be a valid map. Think about setting thresholds near zero in roll and pitch to determine which transformed images are valid for mapping.
Optimizing Time Tip: Moving faster and more efficiently will minimize total time. Think about allowing for a higher maximum velocity and give your rover the brains to not revisit previously mapped areas.
Optimizing % Mapped Tip: Think about ways you can "close" boundaries in your map. If you can do this effectively, then once you close all boundaries your map will be complete!
Optimizing for Finding All Rocks Tip: The rocks always appear near the walls. Think about making your rover a "wall crawler" that explores the environment by always keeping a wall on its left or right. If done correctly, this optimization can help all the aforementioned metrics.

Getting a good result? Post your results in the #udacity_search_map channel in Slack and see how it compares with your fellow students! Check out this livestream discussion of tips, tricks and strategy from some of the top performers so far in this challenge!

2. Collect Samples:

In this project you also have the capability to pick up samples using additional telemetry and command functionality that's available to you. In the drive_rover.py script you'll find the send_pickup() function:

# Define a function to send the "pickup" command 
def send_pickup():
    print("Picking up")
    pickup = {}
    sio.emit(
        "pickup",
        pickup,
        skip_sid=True)
    eventlet.sleep(0)

You can monitor whether or not you are close enough to pick up a rock sample with the Rover.near_sample flag, which will be updated with each new batch of telemetry and will be set to 1 if you are near a rock sample. You should be stopped in order to pick up a rock, so if you are in a state where Rover.near_sample == 1 and Rover.vel == 0 you can set the Rover.pick_up flag to True and the robotic arm will be triggered into action using the following code within the telemetry():

# If in a state where want to pickup a rock send pickup command
if Rover.send_pickup and not Rover.picking_up:
    send_pickup()
    # Reset Rover flags
    Rover.send_pickup = False

Once you have collected all six rock samples, return them to the starting point and you have successfully completed the Udacity Robotics Sample Return Challenge! Share your results with us in the #udacity_search_map channel in Slack!

Questions or issues to resolve before launching into the project? Contact your classroom mentor or ask your question in Knowledge!