Control and Automation of Electric Wheelchair using EEG and AI algorithms
Assistive robots can provide support for disabled people in daily and professional life, thus creating a growing demand for them. Some less severely disabled people can control these devices using joysticks and keyboards but those that are severely disabled cannot communicate with them using convent
2025-06-28 16:30:55 - Adil Khan
Control and Automation of Electric Wheelchair using EEG and AI algorithms
Project Area of Specialization Electrical/Electronic EngineeringProject SummaryAssistive robots can provide support for disabled people in daily and professional life, thus creating a growing demand for them. Some less severely disabled people can control these devices using joysticks and keyboards but those that are severely disabled cannot communicate with them using conventional means. In those cases, special techniques like eye-tracking, sip-and-puff and Brain Computer interface (BCI) are used. There are many techniques available for BCI, but EEG(Electroencephalography) is the most popular technique to implement BCI because it is non-invasive and low cost. An EEG based wheelchair is a wheelchair which uses EEG based Brain computer interface (BCI) to receive human commands. This wheelchair is not only limited to manual-control, but it incorporates aspects of automation and interaction with environment using image processing and artificial intelligence. We are going to develop a wheelchair system which will consist of two cameras and a controller. We will take live stream video from the cameras and using TensorFlow-lite implemented in ImageAI library, alongside OpenCV for image acquisition and object detection, which will all be implemented in Python. An obstacle avoidance and path tracing algorithm will be implemented, which will make the control of electric wheelchair easy, thereby reducing the fatigue on the user experienced due to total manual control. Due to less hardware requirements the cost will also decrease and so will the hardware complexity, thus reducing the risk of hardware failure, which can make it more accessible to middle class people.
- To classify and convert the EEG signals into commands and develop an appropriate algorithm to do so.
- To implement manual control of Electric Wheelchair using EEG signals.
- The design and developement of suitable object detection algorithm for our application.
- To develop a proper pattern recognition algorithm in order to move the Electric Wheelchair in a safe outdoor environment.
- To develope an algorithm for moving the Electric Wheelchair in an indoor environment by avoiding obstacles autonomously.
- To add security features such as live location sharing and battery notifications on mobile app.
- To make a smart decision-making algorithm which is safe and relatively fast.
The EEG(Electroencephalography) command before being used will be pre-processed and classified according to a linear classifier, and then using ERD/ERS (Event related Desynchronization/Synchronization), we will assign proper commands to the corresponding control signals. We will then test the manual control algorithm on wheelchair, and after successful implementation we will move on to the AI (Artificial Intelligence) part of the project. We will use the cameras mounted on the wheel chair to take a live feed , and then using TensorFlow-lite, ImageAI, OpenCV and YOLO(You only live once) as the backend, we will develop an algorithm and train a model, to take decisions based on EEG commands and Object Detection. We will then develop a path tracing algorithm which only requires the user to give the general direction and our algorithm will automatically trace the path by avoiding the obstacles.
Signal acquisition from brain activity using BCI (Brain Computer Interface) can be done through either invasive or non-invasive means, where people almost always prefer non-invasive. In BCIs we used EEG (Electroencephalography) due to its low cost and convenient use. The proposed method since it uses cameras, is of low cost and less bulky compared to other such projects. The use of only cameras as opposed to an array of other sensors, can reduce the cost of repairs as well. Cameras also offer the advantage of allowing the machine to recognize objects and make it more interactive. Using object recognition, we can also increase the safety of the user as it can also detect short abrupt changes in elevations (small pit falls etc.) as opposed to using many ultra-sonic sensors to achieve the same effect. Due to its software heavy design, it can also be upgraded quite easily, as a quick firmware upgrade can add new features and remove any existing bugs. We can also implement facial recognition to detect familiar people. It also reduces the stress on the user due to its semi-automatic nature. The user will also have the option of both fully manual and autonomous control. The manual control will provide more precise control.
Technical Details of Final DeliverableThis project includes a Wheelchair, which will have two 24V 250W DC motors with maximum loading weight of 100Kg excluding its own weight, with maximum climbing angle of less than 10 degrees. A 24V 20AH battery will be used which will give us the approximate range of 10-20km depending on use case. We will use Jetson Nano Developer Kit as the main control unit, two Raspberry Pi Camera Module V2 with 8 Megapixel image resolution and 1080p video resolution will be mounted. An Emotiv Insight (wearable EEG device) is used to acquire the EEG brain activity data. A 10W speaker will be used for more interactivity. The Jetson Nano will have linux4tegra OS installed on them alongside python and its libraries, namely, TensorFlow-lite, ImageAI, NumPy, matplotlib, and OpenCV for python.
Final Deliverable of the Project HW/SW integrated systemCore Industry MedicalOther Industries Transportation Core Technology Artificial Intelligence(AI)Other Technologies RoboticsSustainable Development Goals Industry, Innovation and Infrastructure, Reduced InequalityRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Emotiv insight | Equipment | 1 | 35000 | 35000 |
| Jetson Nano Developer kit with necessary accessories | Equipment | 1 | 25000 | 25000 |
| 24V 20AH Intelligent Fast Battery Charger | Equipment | 1 | 3000 | 3000 |
| Lead Acid Battery 24V 20AH | Equipment | 1 | 7000 | 7000 |
| Travelling Expenses | Miscellaneous | 3 | 1000 | 3000 |
| Printing | Miscellaneous | 10 | 200 | 2000 |
| Wires and connectors | Miscellaneous | 10 | 100 | 1000 |
| Labour | Miscellaneous | 1 | 1000 | 1000 |
| PowerAdapter for jetson nano | Miscellaneous | 1 | 1000 | 1000 |
| Glue gun | Miscellaneous | 1 | 900 | 900 |
| Heat sinks | Miscellaneous | 5 | 20 | 100 |
| Hot glue gun sticks | Miscellaneous | 10 | 40 | 400 |
| Soldering wire 400g | Miscellaneous | 1 | 600 | 600 |