Self-supervised learning for simultaneous localization and mapping (SLAM) of a diverless car
Environment map information is required for the precision agriculture so that the agricultural autonomous machinery can navigate, plan its path, and can perform plantation supervision. The agricultural field is very complex, unstructured, and dynamic environment in which in order to perform autonomo
2025-06-28 16:34:57 - Adil Khan
Self-supervised learning for simultaneous localization and mapping (SLAM) of a diverless car
Project Area of Specialization RoboticsProject SummaryEnvironment map information is required for the precision agriculture so that the agricultural autonomous machinery can navigate, plan its path, and can perform plantation supervision. The agricultural field is very complex, unstructured, and dynamic environment in which in order to perform autonomously precision agricultural operations, the agricultural machinery should have clear information of its location in the field and estimation of path to be followed to reach next location. The SLAM algorithm will implemented on an unmanned non-holonomic car-like mobile robot. The map of the environment is based on the detection of surrounding environment salient information using monocular vision system.
Simultaneous localization and mapping (SLAM) is a process in which a map of an environment is built using available sensory modalities and at the same time, this map is used to localize i.e. finding position/orientation of a car-like mobile robot. In SLAM, both the trajectory of the platform and the location of all landmarks are estimated online without the need for any prior knowledge of location.
The machine learning techniques have led to computationally efficient mapping algorithms. Today, landmarks visual recognition systems are still rarely employed in car-like agricultural mobile robot applications. In this project, an autonomous self-supervised driving platform will be developed for the tasks of optical flow, visual odometry/SLAM and 3D object/landmark detection.
Project Objectives-
The objectives of this project are to develop a car-like mobile robot for localizing a mobile robot by using the salient object tracking approach in an unknown, complex and visually cluttered environment and to develop visual odometry using the real-time input video streams acquired from the on-board installed low cost cameras.
-
Another objective is to develop a computationally efficient mechanism for dynamic mapping of more than 200k 3D object annotations captured in cluttered scenarios to develop dynamic map for simultaneously with localization.
To perform visual SLAM, the first task is to localize an autonomous car-like agricultural mobile robot in an unknown environment. To achieve this task, visual odometry will be performed for the detection, extraction and tracking of salient objects with the help of real-time input video streams. Visual odometry will be performed using PC/BC neural network. The neural network will be at first trained for the recognition of various salient vegetation scenes. The network will get training images from that environment with which it will be trained. After training, the robot will operate autonomously by tracking salient objects in sequence of images and then will determine its position and distance traveled relative to the position of salient objects. Using this process the car-like robot will simultaneously build dynamic map of that unknown environment. This whole process is shown in figure given below:
%20of%20a%20diverless%20car%20_1582926400.png)
The SLAM is an important process of an unmanned non-holonomic car-like agricultural mobile robot for precision agricultural applications. Since the agriculture is a major industry of Pakistan supporting the GDP of economy through agricultural exports. The involvement of agricultural robotics technology for precision agriculture will add extended benefits for the export of agricultural goods and will be a major breakthrough for Pakistan’s agricultural industry in order to increase its yield in parallel to the rising demands of agricultural products.
Technical Details of Final DeliverableAn unmanned non-holonomic car-like agricultural mobile robot equipped with a complete system for SLAM will be available as deliverable. The developed robot will be fully functional to be utilized for precision agricultural tasks.
Final Deliverable of the Project Hardware SystemCore Industry AgricultureOther IndustriesCore Technology RoboticsOther TechnologiesSustainable Development GoalsRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Arduino | Equipment | 4 | 700 | 2800 |
| DC Motor | Equipment | 5 | 1500 | 7500 |
| Tyre | Equipment | 4 | 600 | 2400 |
| Ras-Pi-Pi4 | Equipment | 2 | 12000 | 24000 |
| Camera | Equipment | 4 | 4500 | 18000 |
| GPS Module | Equipment | 1 | 1190 | 1190 |
| Vero board | Equipment | 5 | 76 | 380 |
| Breadboard | Equipment | 3 | 150 | 450 |
| L298D IC | Equipment | 4 | 220 | 880 |
| Steper Motor | Equipment | 4 | 600 | 2400 |
| Structure Development Labour (Welding and cutting etc) | Miscellaneous | 1 | 4000 | 4000 |
| Battery | Equipment | 1 | 7000 | 7000 |
| Hitec HS-805BB | Equipment | 14 | 200 | 2800 |
| jumper wire | Equipment | 1 | 200 | 200 |
| Printing | Miscellaneous | 1 | 2000 | 2000 |
| Iron bars and plates | Miscellaneous | 4 | 1000 | 4000 |