Team RUDRA System Acceptance Review | University Rover Challenge

Google Cartographer enables SLAM in GPS-denied environments with less than 0.3m drift, reconstructing high-fidelity 3D maps for post-processing. Two world modeling approaches exist: map processing finds frontiers where known meets unknown data, while mission-based modeling tracks specific objectives like radiation sources. Frontiers are detected using hessian transforms on occupancy grids, identifying edges where light-to-dark transitions occur, though computational constraints require maintaining only 20 active frontier candidates.

SLAM is the core technique for onboard state estimation, also called structure from motion. The problem involves a robot starting at position X0, measuring environmental features, moving to new positions, and measuring overlapping and new features. The solution constructs a graph with two edge types: sensor measurements (cameras, laser scanners) and motion estimates (based on applied inputs and time). Solving this optimization problem recovers both feature positions and robot displacements. Effective implementation requires fusing multiple sensors (GPS, laser scanners, IMUs, cameras) with different characteristics and rates using filters to produce state estimates at 100-200 Hz.

The video shows examples of SLAM in GPS-denied environments. In underground mining environments, robots must simultaneously build a map of the mine while localizing themselves within that map. The video demonstrates SLAM integrating GPS measurements with wheel odometry: GPS provides absolute position information but can be lost in areas with poor satellite visibility (like dense forests), while wheel odometry provides continuous position estimates but accumulates errors over time. The video shows that combining both sources helps create a more accurate map and trajectory. The video also shows SLAM in indoor environments where GPS is unavailable, relying on wheel odometry and landmark detection. The video demonstrates that wheel odometry alone can lead to significant errors, especially in rotation estimation.

vSLAM (Visual Simultaneous Localization and Mapping) is a navigation technology that uses cameras and sensors like IMUs to build environmental maps and determine location in GPS-denied areas such as urban canyons, forests, or underground mines; it addresses the drift problem in GPS navigation by enabling loop closure, where the system recognizes previously seen locations to reset accumulated errors and maintain accurate positioning over long distances.

Sensor fusion combines multiple independent observations of the same quantity to improve estimation accuracy. In multi-UAV navigation, each vehicle provides GPS position while also measuring relative distances to teammates. The observation matrix maps the augmented state vector to all available measurements. SLAM (Simultaneous Localization and Mapping) addresses the fundamental challenge of determining a robot's location while simultaneously building a map of unknown environments. In underwater applications where GPS is unavailable, SLAM uses relative measurements to fixed landmarks combined with inertial measurements. Without landmark constraints, position error grows linearly with time due to integration of noisy velocity measurements. Landmark measurements provide absolute references that reset this error growth, enabling accurate long-term navigation in GPS-denied environments.

Scientists use three main techniques for planetary exploration: multispectral and hyperspectral imaging (capturing images at different light wavelengths to identify materials), Raman spectroscopy (using laser light to excite molecules and analyze their spectral signatures), and gas chromatography-mass spectrometry (separating and identifying chemical compounds). Sample return missions like OSIRIS-REx successfully brought asteroid Bennu samples to Earth, where scientists found amino acids—building blocks of life. These methods allow comprehensive analysis of extraterrestrial materials without returning samples to Earth.

Perseverance carries seven scientific instruments: Mastcam-Z (zoomable multispectral imager), SuperCam (remote laser-induced breakdown spectroscopy), PIXL (X-ray fluorescence spectrometer), SHERLOC (Raman and UV-induced fluorescence spectrometer with microscope), MOXIE (oxygen extraction demonstrator), RIMFAX (ground-penetrating radar), and MEDA (environmental monitoring station). These instruments work together to analyze rocks and soil using multiple spectroscopic techniques across different wavelengths to detect organic compounds and characterize sample composition.

The Mars 2020 rover carries seven complementary scientific instruments designed to investigate Mars' habitability, past life, and atmospheric conditions: Mastcam-Z provides zoom-enabled multispectral imaging for terrain navigation; SuperCam uses laser-induced breakdown spectroscopy, Raman spectroscopy, and visible/IR spectroscopy for remote elemental and mineral analysis; PIXL employs X-ray fluorescence for fine-scale elemental mapping; SHERLOC uses UV Raman spectroscopy for detecting organic compounds; MOXIE produces oxygen from CO2 for future human exploration; MEDA monitors weather including dust characteristics; and RIMFAX uses ground-penetrating radar to image subsurface geology to 50 cm resolution. These instruments overlap in their analytical capabilities to ensure robust scientific results and guide sample selection for potential return to Earth.

SuperCam on Perseverance integrates five analytical techniques into a single remote sensing instrument: LIBS for elemental composition, UV Raman for molecular structure, time-resolved Raman for standoff analysis at several meters, visible-infrared reflectance spectroscopy for passive mineralogy at kilometers range, and audio recording for environmental monitoring. Time-resolved Raman uses pulsed lasers (14 mJ at nanosecond duration) with time-gated detection to isolate Raman signals from background light. Magnetic protection systems inherited from Curiosity's ChemCam keep calibration targets clean. This multi-technique approach maximizes scientific return from limited payload mass, demonstrating how integrated instrumentation enables comprehensive characterization of Martian targets through complementary analytical capabilities.

The NASA Mars 2020 rover is equipped with seven advanced scientific instruments designed to search for signs of life on Mars, including a sophisticated camera system for panoramic and stereoscopic imaging, a powerful laser (SuperCam) for analyzing rock composition, a gas analyzer (MEDA) for studying Martian weather conditions, a subsurface radar (RIMFAX) capable of penetrating up to 10 meters into the Martian surface, a X-ray spectrometer (PIXL) for precise elemental analysis, an ultraviolet Raman spectrometer (SHERLOC) for detecting organic compounds and minerals, and a methane sensor (Methane Sensor) for detecting atmospheric methane.

Communication delays severely impact teleoperation performance by preventing operators from maintaining intended paths and limiting maximum speeds. Model-free predictors compensate for these delays using first-order time delay systems with single stability parameters, requiring no knowledge of driver or vehicle dynamics. Human-in-the-loop testing demonstrates that predictors reduce track completion time by 29%, lane keeping error by 46%, and steering control effort by 47% compared to delayed-only conditions. This enables high-speed teleoperation despite significant communication latencies.

Ultra-low latency teleoperation technology enables real-time remote control of machines like robots, drones, and vehicles by using UDP-based protocols that synchronize multiple data streams (video, audio, control inputs) while accounting for clock drift, achieving glass-to-glass latency as low as 4 milliseconds to make distance effectively disappear for applications ranging from cloud gaming to remote surgery.

TWIST achieves remarkable latency reduction from approximately 0.9 seconds to below 0.1 seconds through systematic engineering improvements. The key insight is identifying bottlenecks in the system and addressing them sequentially. The retargeting stage initially consumed most latency, but optimization reduced this significantly. Low latency is critical for teleoperation because operator overcompensation for perceived delay makes tasks slower and more difficult. At latencies below human perception thresholds, operators can perform naturally without compensatory adjustments.

This extensive section presents the complete technical framework for teleoperated driving. It covers teleoperation as infrastructure support for autonomous vehicles, explaining direct versus indirect control methods. The presentation details critical system requirements including vehicle limitations, task prioritization, safety barriers, guaranteed communication capacity, low latencies, high uplink data rates, scalability, and redundancy. The second half introduces managed latency technology based on the IETF L4S standard, explaining how prioritization and network-application feedback loops using ECN bits maintain stable latency for real-time applications including teleoperation, cloud gaming, AR/VR, and video conferencing.

This video explains three UHF antenna types for long-range radio systems: the half-wave dipole (most efficient at ~100%, balanced, but large), the quarter-wave monopole (half the size but creates ground plane interference with video transmitters, limiting range), and the coil-loaded dipole (compromise between size and efficiency at ~75%, with narrow bandwidth). Construction methods involve cutting RG316 coaxial cable, preparing wire elements, and soldering connections, with performance verified using a Vector Network Analyzer.