|
|
 |
|
 |
johnrennceo
Grünschnabel
Level: 1 
Erfahrungspunkte: 1
Nächster Level: 10
|
 |
|
| New Algorithm Boosts Efficiency of Lidar-Based Delivery Robots |
  |
Autonomous delivery robots from companies like Starship Technologies and Kiwibot are increasingly navigating city streets and neighborhoods. These robots rely on a combination of sensors and software to understand and move through their environments.
A key technology in their navigation systems is lidar, which uses light pulses to measure distances and create detailed maps of surroundings through a process called simultaneous localization and mapping, or SLAM. While effective, this approach consumes significant memory and computing power, limiting how far and efficiently robots can operate.
"Over time, these systems can accumulate 10 to 20 gigabytes of data," says Zihao Dong, a doctoral student at Northeastern University. "Managing that much information becomes a heavy computational burden."
To overcome these challenges, Dong has developed a new 3D mapping technique that significantly reduces the computational load required for robotic navigation. Working under the guidance of Professor Michael Everett from Northeastern's Department of Electrical and Computer Engineering, Dong introduced a method that cuts memory use by up to 57 percent compared to leading alternatives. Their findings are published on the arXiv preprint server.
The algorithm, called Deep Feature Assisted Lidar Inertial Odometry and Mapping (DFLIOM), builds on an existing method known as Direct LiDAR Inertial Odometry and Mapping (DLIOM). Both rely on lidar sensors and inertial measurement units to construct 3D maps, but DFLIOM enhances the process by using a more efficient way of interpreting environmental data. In some cases, it even improves mapping accuracy.
 https://github.com/BrianTMN/Rainbow-Six-...unlimited-money
https://github.com/JasonBTD/Temple-Run-2...ns-and-gems-MOD
https://github.com/KevinBMT/50-Tiny-Room...s-and-coins-MOD
https://github.com/DavidFNT/Grand-Gangst...mited-money-MOD
https://github.com/AndrewGNT/Happy-Diner...ey-and-gems-MOD
https://github.com/StephenGNA/Realms-of-...ited-everything
https://github.com/EricBNT/Run-Goddess-M...-money-and-gems
https://github.com/AlexanderPGN/State-Co...ey-and-gems-MOD
https://github.com/KyleNBO/Spider-Fighte...mited-money-MOD
https://github.com/ScottBND/Flex-City-un...ey-and-gold-MOD
https://github.com/TravisMBR/Agent-Hunt-unlimited-money-MOD
https://github.com/ShaneGNC/War-Machines...nd-diamonds-MOD
https://github.com/TimothyKBT/Standoff-2-unlimited-gold-MOD
https://github.com/ChrisBNC/Zombie-Apoca...ey-and-gold-MOD
https://github.com/GregTNB/Critical-Stri...mited-money-MOD
https://github.com/JoelKNV/Space-Survivo...ey-and-gems-MOD
https://github.com/TroyMBC/World-War-Her...imited-gold-MOD
https://github.com/DennisKMB/Once-Human-...imited-Crystgin
https://github.com/DevonMNB/Suzy-s-Resta...mited-money-MOD
https://github.com/DustinBRM/FPS-Strike-...mited-money-MOD
https://github.com/KevinHSN/Grunt-Rush-B...-money-and-gems
https://github.com/DaveALG/Top-Heroes-Ki...ited-everything
https://github.com/jasonRBL/Archers-Onli...ey-and-gems-MOD
"This research pushes back against the assumption that more data automatically means better performance," says Everett. "Sensor companies often promote devices that deliver increasingly dense data, but more data can actually make it harder for algorithms to keep up."
Dong and Everett's work focuses on designing smarter algorithms that extract only the most essential information, allowing robots to map their environments more efficiently without compromising performance.
The new approach was tested using Northeastern’s Agile X Scout Mini robot, equipped with an autonomy kit including an Ouster lidar, battery pack, and Intel NUC mini PC. The robot successfully created 3D maps of several campus locations, including Centennial Common, Egan Crossing, and Shillman Hall.
This advancement could lead to more capable, longer-lasting delivery robots that can operate effectively without overwhelming their onboard systems—an important step forward in the quest for scalable urban automation.
|
|
0
Gestern, 22:36 |
|
|
 |
|
 |
|
|
Besucher seit 04.04.2003 gesamt: 12.784.958 | Besucher heute: 3018 | Besucher gestern: 10.969 |
|
|
|
