RESEARCH

SMAV

IMG-20230913-WA0012
photo_2023-09-13_19-13-51
IMG-20230918-WA0002

SMAV

  1.  The Shawplateless configuration is an innovative drone project that eliminates traditional landing gear for a streamlined, plateless design.
  2. Advantages include enhanced aerodynamics, reduced weight, increased flight efficiency, and extended operational range.
  3. The absence of landing gear reduces the risk of damage during takeoff and landing, especially in constrained or rugged environments.
  4. Challenges include ensuring structural integrity, developing precise control algorithms and sensor integration, and adapting regulatory frameworks.
  5. Despite challenges, the project aims to revolutionize drone technology with new possibilities through dedicated research and collaboration.

VTP

2
Screenshot 2023-09-23 at 23.45.01

VTP

  1. The Visual Trajectory Planning project focuses on advancing visual-planning knowledge using Image Analysis and non-conventional Computer Vision techniques.
  2. It implements a feature for recognizing gates and computing suitable drone trajectories using neural networks.
  3. The task is divided into two parts: gate identification with a neural network for distance measurement, and trajectory identification with a second neural network for optimal path planning.
  4. This feature is crucial for ensuring the drone can safely navigate both internal and external environments.
  5. The initial hardware implementation will utilize a DJI Tello drone.

LTS

LTS_1
MAT_1
MAT_2

LTS

  1. The LTS (Long Term SLAM) project explores various approaches to solve or improve the SLAM (Simultaneous Localization and Mapping) problem.
  2. It addresses multiple issues related to SLAM in both hardware and software fields.
  3. The project involves studying and testing different setups to build a comprehensive database of SLAM-related information.
  4. It aims to develop efficient localization and mapping strategies, focusing on sensor suite choices, sensor fusion algorithms, and advanced techniques like Deep Learning.
  5. All solutions are tested and implemented in hardware to evaluate and compare their effectiveness.

RLBoTs

RLB_1
RLB_2
RLB_3

RLBots

  1. The project develops Reinforcement Learning (RL) algorithms to enhance robot autonomy.
  2. RL focuses on balancing exploration and exploitation without needing labeled input/output pairs.
  3. These algorithms have significant generalization potential due to their learning paradigm.
  4. Modeling the simulation environment and training the algorithms are challenging and active research areas.
  5. The current goal is to develop a Deep RL Agent for obstacle avoidance in unknown environments, progressing towards full autonomous movement and planning.

AMAV​

AMAV_1

AMAV

  1. Micro Aerial Vehicles (MAVs) are small UAVs known for their ease of transport, high maneuverability, and flexibility in tight spaces.
  2. Advantages of MAVs include rapid deployment, easy swarm implementation, low-risk of damage, and expendability.
  3. The research focuses on developing MAVs with extended autonomy and flight duration capable of autonomously following an object using deep learning algorithms.
  4. The project aims to achieve full on-board autonomous control, covering sensor data acquisition, processing, decision making, and flight control.
  5. Key technologies investigated include miniaturized high-performance cameras, compact avionics, high specific power supplies, telecom, and actuators.

JARIM

JARIM - Technology Demonstrator
JARIM - Technology Demonstrator
JARIM - Technology Demonstrator
JARIM - Technology Demonstrator

JARIM

  1. JARIM (Just A Rather Intelligent Machine) is a modified RC car designed to autonomously drive within a circuit using a mono camera for image acquisition.
  2. The car’s decision-making for throttle and steering is handled by a machine learning algorithm, developed via reinforcement learning in a simulated environment.
  3. An image segmentation algorithm was created and trained to label the circuit lines using a dataset of circuit images.
  4. The car’s actuators were modified to be controlled by an onboard computer, and a new electric and power system was integrated.
  5. For safety, both wired and wireless E-stops were implemented, and the Robot Operating System (ROS) was used for managing camera data, decision making, and actuator control.