This video demonstrates a complete autonomous UAV mission where Team Arrow's ASCEND drone executes a full mission cycle including autonomous takeoff, lawn mower survey pattern navigation using vision-based feature tracking, precise landing using April tags for position estimation, autonomous docking and charging via pogo pin connectors, and post-mission data processing including image stitching, feature detection, and validation. The system uses optical flow for stable flight, April tags for reference frame establishment, and a two-phase landing approach transitioning from corner tags to a center tag for precision docking.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
ISRO Robotics Challenge 2026 – Elimination Tasks Video | Team Arrow, Nirma University | TID: 11268
Added:Ascend arms and takes off autonomously climbing to its server altitude of 3.5 m while maintaining stable flight using optical flow and on board sensing.
After reaching server altitude, the drone visually detects four April tags on the landing pad and precisely centers itself above the base station. This location becomes 00 coordinate establishing the reference frame throughout the mission.
The drone then proceeds to follow a lawn mower survey pattern to ensure complete coverage of the arena. The waypoints for this path are computed based on the camera's field of view, the dimensions of the arena, the survey altitude, the required horizontal and vertical image overlap, and the base station position in respect to the arena.
As the vehicle traverses the survey path, it continuously estimates its local position using vision-based navigation.
Distinct ground features are tracked across consecutive video frames, and the resulting feature drift is used to determine the drone's displacement relative to the mission origin. This enables accurate navigation and positioning.
At each predefined capture point, the drone briefly stabilizes and acquires an image of the survey area. The overlapping image ensures complete arena coverage while providing redundancy for target localization and validation.
After completing the survey mission, Ascent autonomously returns to the origin of the reference frame and begins the landing sequence.
The vehicle first re-acquires the four April tags located at the corners of the base station and precisely aligns itself with the pad center.
Once the alignment is confirmed, a descent is initiated in two phases.
During the first phase of landing, the drone uses the four corner April tags for position and orientation estimation while descending towards the base station.
As the vehicle approaches the pad, the corner April tags gradually move out of camera's field of view.
At approximately 1 m above the base station, the system transitions to the second phase of landing. In this phase, Ascend relies on the smaller April tag positioned at the center of the pad. The drone continuously tracks this tag, refining its position and yaw alignment to maintain precise centering throughout the remainder of descent.
Using the center April tag as the final visual reference, Ascend performs a controlled precision landing onto the docking station. Once the drone is successfully docked, the pogo pin connectors come in contact with the brass plate on the autonomous charging sequence begins.
When the landing gear bridges two pads, a sensing circuit detects a contact and sends a signal to the microcontroller through an ADC. The controller then determines the required polarity and activates the appropriate P-MOS and N-MOS switches, configuring the contact pads as positive and negative terminals.
Power is then routed through the charging matrix and the battery begins charging automatically, enabling a fully autonomous charging cycle without any human intervention.
The images that are captured and saved during the mission are transferred to the base station computer for testing and validation.
All the 19 images are received and now they are being sent for map generation.
The images are stitched together according to their features for the map generation. This map is further used for the localization and detection of the features.
Once the map is generated, it is passed to the detection algorithm for the identification of the features. The pixels of the detected features are then converted to the real world coordinates and are showcased.
The detected feature images are then converted to the low resolution images and are validated with the seed images that are already fed into the base station computer.
>> Hey.
Related Videos

Setting up a curved screen with Immersive Calibration Pro 4 and multiple cameras (P3D v4)
FlyerOneZero
23K views•2019-07-21

Robot Learning with Sparsity and Scarcity
allenai
379 views•2025-10-14

Jorge Mendez-Mendez: Unlocking Lifelong Robot Learning With Modularity (2023-10-05)
umassmlfl
237 views•2024-01-06

Northwestern’s MS in Robotics: Student Robotics Projects, 2023
NorthwesternEngineering
1K views•2024-05-31

"Perfect" Turns: Turning by the Gyro - FIRST LEGO League (FLL) SPIKE Prime + EV3 RePlay Programming
ZacharyTrautwein
94K views•2020-10-02

Gorkem Secer: TSLIP-based Deadbeat Running Control of Bipedal Robot ATRIAS
DynamicWalking-wv6qm
298 views•2018-06-22

Self-Driving Cars Need Lessons On Human Drivers | Maddie About Science
skunkbear
26K views•2018-08-21

Milrem Robotics’ THeMIS UGVs used in a live-fire manned-unmanned teaming exercise
MilremRobotics
99K views•2021-05-20
Trending

NOLAN WELLS AUTOPSY: DEEP TISSUE DISCOLORATION BACK OF HEAD, NECK BONE MISSING
nancygrace
561K views•2026-07-22

Big Tech's Biggest Gamble Is Finally Falling Apart
houseofel-ai
79K views•2026-07-22

we're almost finished the house (ep.125)
JennaPhipps
347K views•2026-07-22

We flooded a field - the results blew our minds
MossyEarth
73K views•2026-07-22