Teighan Miller - Bootcamp #142

teighanmiller · 2025-01-13T22:00:04Z

No description provided.

TongguangZhang

reviewed

TongguangZhang · 2025-01-14T01:45:09Z

modules/bootcamp/decision_simple_waypoint.py

@@ -68,6 +66,25 @@ def run(
        # ↓ BOOTCAMPERS MODIFY BELOW THIS COMMENT ↓
        # ============

+        # If the drone is halted and not at the destination, move the drone to destination
+        if report.status == drone_status.DroneStatus.HALTED and report.destination != self.waypoint:


use the acceptance radius, not != (float operations are a little funky, so you can't do a direct comparison like this)

TongguangZhang · 2025-01-14T01:45:46Z

modules/bootcamp/decision_simple_waypoint.py

+            command = commands.Command.create_set_relative_destination_command(
+                self.waypoint.location_x, self.waypoint.location_y
+            )


note that it's relative distance, not absolute, so you can't just use the coords of the waypoint

TongguangZhang · 2025-01-14T01:46:22Z

modules/bootcamp/decision_simple_waypoint.py

+        elif (
+            report.status == drone_status.DroneStatus.HALTED and report.destination == self.waypoint
+        ):


same as above - use acceptance radius instead of ==

TongguangZhang · 2025-01-14T01:47:39Z

modules/bootcamp/decision_simple_waypoint.py

+        elif (
+            report.status == drone_status.DroneStatus.HALTED and report.destination == self.waypoint
+        ):


it seems like you make a redundant calculation here - you can nest the if statements, which lets u do if else

if halted: if at location: ... else: ...

TongguangZhang · 2025-01-14T01:48:21Z

modules/bootcamp/decision_simple_waypoint.py

+        if (
+            report.status == drone_status.DroneStatus.LANDED
+            and (report.position - report.destination) > self.acceptance_radius
+        ):
+            command = commands.Command.create_set_relative_destination_command(
+                self.waypoint.location_x, self.waypoint.location_y
+            )


it doesn't make sense to land the drone and then take off again if it's in the wrong spot - just make sure it's in the right spot before landing

TongguangZhang · 2025-01-14T01:55:34Z

modules/bootcamp/detect_landing_pad.py


        # Detach the xyxy boxes to make a copy,
        # move the copy into CPU space,
        # and convert to a numpy array
-        boxes_cpu = ...
+        boxes_cpu = prediction.boxes.xyxy.cpu().numpy()


make sure to call detach(), this removes it from the computational graph, which makes it stop tracking the gradients
basically, during training, gradients are tracked on the tensors for back propagation, but since we're only using inference, we don't need the gradients, and they take up a lot of memory, so we should stop tracking them

TongguangZhang · 2025-01-14T01:55:56Z

modules/bootcamp/detect_landing_pad.py


        # Loop over the boxes list and create a list of bounding boxes
        bounding_boxes = []
+        for i in range(0, boxes_cpu.shape[0]):
+            new_box = bounding_box.BoundingBox.create(boxes_xyxy[i])


better practice to do result, new_box = ...

TongguangZhang · 2025-01-14T01:56:15Z

modules/bootcamp/detect_landing_pad.py

+            if new_box[0]:
+                bounding_boxes += new_box


instead of skipping failed boxes, we acc wanna return an empty list - in data science, when part of a data point is corrupted, we skip the whole data point instead of returning partial data, in this case, 1 image is 1 data point, and the boxes are part of the image, so we skip the whole image if any of the boxes fail

TongguangZhang · 2025-01-14T01:57:12Z

modules/bootcamp/detect_landing_pad.py

+        for i in range(0, boxes_cpu.shape[0]):
+            new_box = bounding_box.BoundingBox.create(boxes_xyxy[i])
+            if new_box[0]:
+                bounding_boxes += new_box


use .append for lists - also, do you need to add the success flag to the bounding box list instead of just the box

TongguangZhang · 2025-01-14T01:57:40Z

modules/bootcamp/detect_landing_pad.py

+        if not bounding_boxes:
+            return [], image_annotated


this can be deleted if you don't add the success flags

…ptance radius

maxlou05

Reviewed

maxlou05 · 2025-01-18T16:25:53Z

modules/bootcamp/decision_simple_waypoint.py

@@ -68,6 +66,16 @@ def run(
        # ↓ BOOTCAMPERS MODIFY BELOW THIS COMMENT ↓
        # ============

+        # If the drone is halted and not at the destination, move the drone to destination
+        if report.status == drone_status.DroneStatus.HALTED:
+            if (report.destination - self.waypoint) < self.acceptance_radius:


Please use Euclidean distance and not this. I'm not even sure how this comparison works, as you are comparing a vector (in this case a class with 2 ints) with a scalar (1 int).

maxlou05 · 2025-01-18T16:30:10Z

modules/bootcamp/decision_waypoint_landing_pads.py

@@ -38,6 +38,8 @@ def __init__(self, waypoint: location.Location, acceptance_radius: float) -> Non
        # ============

        # Add your own
+        self.target = None
+        self.tol = 0.0001


The tolerance is given to you as acceptance_radius

maxlou05 · 2025-01-18T16:31:37Z

modules/bootcamp/decision_simple_waypoint.py

+        if report.status == drone_status.DroneStatus.HALTED:
+            if (report.destination - self.waypoint) < self.acceptance_radius:
+                command = commands.Command.create_set_relative_destination_command(
+                    abs(self.waypoint.location_x), abs(self.waypoint.location_y)


It is a relative distance command, so your own position matters. Also, don't use abs because (-1, -1) is a valid location and is different from (1, 1)

maxlou05 · 2025-01-18T16:31:55Z

modules/bootcamp/decision_waypoint_landing_pads.py

+        if report.status == drone_status.DroneStatus.HALTED and report.destination != self.target:
+
+            command = commands.Command.create_set_relative_destination_command(
+                self.target.location_x, self.target.location_y


Relative distance command, your position matters

maxlou05 · 2025-01-18T16:33:58Z

modules/bootcamp/decision_waypoint_landing_pads.py

+        if (
+            report.status == drone_status.DroneStatus.LANDED
+            and (report.position - report.destination) > self.acceptance_radius
+        ):
+            command = commands.Command.create_set_relative_destination_command(
+                self.target.location_x, self.target.location_y
+            )
+


Instead of doing this extra case, can you just add to the condition of the elif above? This doesn't really work right now because the drone doesn't accept move commands after it lands, so incorporate it into the one above

maxlou05 · 2025-01-18T16:35:36Z

modules/bootcamp/decision_waypoint_landing_pads.py

+        closest destination. This location.Location object is returned.
+
+        Args:
+            current_location (location.Location): Location the drone is currently at (0, 0)


It starts at (0,0), but you cannot assume the drone doesn't move (you don't know where the drone is when you call this function). However, it doesn't really matter so it's fine. Returning to (0,0) is ok if there are no landing pads, great job for thinking about this case! (You do not need to change anything here)

maxlou05 · 2025-01-18T16:36:49Z

modules/bootcamp/detect_landing_pad.py


        # Loop over the boxes list and create a list of bounding boxes
        bounding_boxes = []
+        for i in range(0, boxes_cpu.shape[0]):
+            result, new_box = bounding_box.BoundingBox.create(boxes_xyxy[i])


Please use boxes_cpu here

… and euclidean distance

maxlou05

Reviewed. Also check discord about the distance thing

maxlou05 · 2025-01-18T21:30:57Z

modules/bootcamp/decision_waypoint_landing_pads.py

+        if self.target is None and report.status is drone_status.DroneStatus.HALTED:
+            self.target = self.find_closest_landing_pad(report.position, landing_pad_locations)


For task 4, you need to fly to the waypoint first (you can think of it as target1), before you can fly to the nearest landing form the waypoint (target2). So in total you need to go to 2 things, but you only go to 1 (the landing pad)

maxlou05

Conditionally approved! I would advise to be careful when using is, because I think that is just compares pointers and not compares objects (does not invoke the object's eq()). I think it works in your case since most object assignments are just references. This is just an FYI for future.

Teighan Miller added 7 commits January 8, 2025 10:40

Removed GPU requirement

218985e

Completed detect_landing_pad.py

bae10d8

Changed FPS for simulator

c0f0d5d

Completed task UWARG#3

0114e4f

Ran Pylint, black, and flake8

0274fa7

Added logic for task UWARG#4

bcc4438

formatting corrections

7384396

TongguangZhang reviewed Jan 14, 2025

View reviewed changes

Teighan Miller added 5 commits January 14, 2025 14:02

fixed decision_simple_waypoint.py: removed redundancies and used acce…

72e0c10

…ptance radius

Fixes based on feedback

d71d917

Fixed suggestions

27f52af

changed distance funciton name

e53be2d

ran formatters

2f9be6a

maxlou05 reviewed Jan 18, 2025

View reviewed changes

Teighan Miller added 2 commits January 18, 2025 16:11

Changed simple_waypoint and waypoint_landing to use acceptance radius…

cd86453

… and euclidean distance

ran formatters

6ec5388

maxlou05 reviewed Jan 18, 2025

View reviewed changes

Teighan Miller added 5 commits January 18, 2025 16:32

Changed distances to euclidean distance

2499f48

used boxes_cpu

235dadd

removed redundant check for bounding_boxes

e60e5bb

completed task 4

f38c794

ran formatters

bf58708

maxlou05 reviewed Jan 19, 2025

View reviewed changes

maxlou05 closed this Jan 24, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Teighan Miller - Bootcamp #142

Teighan Miller - Bootcamp #142

teighanmiller commented Jan 13, 2025

TongguangZhang left a comment

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

TongguangZhang Jan 14, 2025

maxlou05 left a comment

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 Jan 18, 2025

maxlou05 left a comment

maxlou05 Jan 18, 2025

maxlou05 left a comment

		if self.target is None and report.status is drone_status.DroneStatus.HALTED:
		self.target = self.find_closest_landing_pad(report.position, landing_pad_locations)

Teighan Miller - Bootcamp #142

Teighan Miller - Bootcamp #142

Conversation

teighanmiller commented Jan 13, 2025

TongguangZhang left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment