• Autonomous



Cognitive and Autonomous systems

Uurmi's expertise in artificial intelligence, computer vision and multi sensor data fusion gives it a unique capability of designing cognitive and autonomous system for the battlefield. These could be autonomous ADV's, self-navigating aerial or underwater vehicles, or small robots for specialized operational tasks.

Autonomous Aerial Vehicles:

Autonomous Aerial Vehicles (AAV) has gained a lot of attention in the last few years. Especially quad-copters (micro aerial vehicle: MAV) are becoming smaller, stable, and has potential applications in defense, surveillance, and search and rescue operations. At Uurmi, we are focused on developing AAV's with sophisticated capabilities.

Uurmi’s AAV's are equipped with Global Positioning System (GPS) along with Inertial Measurement Unit (IMU) for navigation from one point to another. In a GPS denied environment (due to jamming of the signal) other sensors such as Radars, Lidar and video camera are used to localize the AAV's position. We have developed algorithm such as Point Mass Filter that uses reference Digital Terrain Elevation Data (DTED) and the measured elevation based on radar and barometer to localize the AAV. In parallel, we are developing vision based perception, navigation and control of aerial vehicle that can travel through windows, corridors, stairs and other complex environments. We use state-of-art machine learning, image processing and computer vision techniques to achieve this goal. As a first step toward autonomous navigation, Uurmi has developed robust vision and control algorithms to automatically land the MAV on a Landing Pad. The algorithm has been ported to platform such as Raspberry Pi and could be easily ported on to other embedded systems.

Autonomous Ground Vehicles:

Uurmi is developing a low cost autonomous car solution. For example most of the autonomous cars use a very expensive laser range finder for detection obstacles and free space before the vehicle but we have developed algorithms using a stereo camera to do the same at a fraction of cost. In this endeavor of making an autonomous car the first step was to control the car electronically i.e to develop a drive by wire system.

We motorized the brake and accelerator pedal, we tapped into the motor that drives the power steering module inside a car and fitted our own electronic actuation modules. We fitted encoders to the motors that drive the pedals and steering, put our own PID servo control electronics and were able to drive the car from a computer. We also fitted a joystick module such that the car can be manually controlled with a wireless remote. The challenge in setting up the PID electronics module was to adjust the response of the car. So we did rigorous manual testing and adjusted the proportional constant and differential constants in the PID to make the car work on any surface. Be it a parking lot or sand or over black tar road. Each surface offers different friction and each of them requires different driving torque from the power steering motor.

Autonomous AFVs:

Sensor suite

The sensor suit would be a combination of various navigation and positioning sensors like ceo, IR cameras, liDAR for navigation, target detection and tracking and GPS/GlONASS+IMU based INS for positioning etc.

Learning and Adaptation

The algorithm suite is highly intelligent and adaptive in nature. The module learns response times and varying inputs over a certain time frame and intelligently adapts itself to give out the most optimum calculations in the battlefield.

Multi Sensor Data Fusion

The algorithm suite module collects data from various on board sensors like positional sensors, navigation sensors etc. and combines the intelligence gathered to give out a multi it into a layered output which can be easily understood by the user and can be an input for the counter measure strategies.

Route Planning & Navigation

Taking the overall path of the vehicle into consideration from the GPS data a global route is planned. later, the inbuilt algorithms specific to route planning take care of the immediate route plan and obstacle avoidance of the vehicle using laser and IMU data.

Vehicle control

Embedded power controls for steering and vehicle's accelerator pedals are used to make the vehicle move in desired direction and at a desired velocity.

Collision Avoidance

The 3D depth data from the stereo cameras and 2D depth data from the LiDAR sensors are used to find the immediate navigable area taking into consideration the obstacles and then the safer route is planned automatically.