Project Duration : July 2018 – Dec 2018 | August 2019 – Present
About : Sparse Map unlike Dense Map, stores only few good features that can be used for visual localization. It eliminates the dynamic features upon multiple loop closures, and stores only static features. It becomes computationally very efficient to work with Sparse Map instead of Dense Map for Real Time Localization.
Current Status : Testing is going on the new mathematical model developed by me, with is suppose to remove parallel placement constraints and reduce average error by more than 30 percent.
Project Inspiration :
1. The least count of our test prototype’s wheel encoder is 1-kmph which was very poor for an autonomous car, and caused poor performance of controls and planner. Thus, there was a need of a more accurate wheel encoder.
2. Since, there was no extra space for adding another optical-based encoder, there was a need of something that can be attached externally. That’s why there was a need of non-intrusive and more accurate wheel encoder.
Goal : Develop an end to end non-intrusive product, that can be attached with any ground vehicle bot or prototype externally. Improved accuracy over existing optical based encoder.
INIOS Ver.1.0 [Intelligent Non-Intrusive Inertial Odometry Sensor]
IMU : Polulu Alt-IMU10-v5
nRF24L01+
Arduino Nano
Bug-Boost Converter
Charging Module
12V Rechargeable Battery
design and development of a non-intrusive inertial sensor used for determining the odometry of a ground vehicle. The sensors are meant to be placed on the rim of the wheels of a ground vehicle and determine the angular velocity of wheels and the angular velocity of the vehicle about its center axis perpendicular to the ground plane. The mathematical model incorporates the electronic gyroscope data of left and right wheel sensor and the relationship between the angular velocity about the central axis of wheels on either side while turning, to calculate the turning angular velocity of car and angular velocities of wheels which in turn determines the linear velocity of the car. These sensors can be used to replace the conventional optical or hall-effect based speed sensors which are difficult to be used with existing models of the ground vehicle as they require specific placement and specially designed mounts. The primary advantage of this sensor is that it can be placed at any position on the wheel, at any angle with respect to the wheel plane. The determined parameters do not depend on the placement of the sensor on the wheel in any way, as it only requires the resultant of angular velocities on each tire. Multiple noises and bias filters such as Kalman Filter for bias reduction and Extended Kalman Filter and Madgwick Filter among others for noise reduction have been tested on this model and the obtained results have been compared and analyzed in this paper with respect to the quality of the output, computation time and resultant lag. Different modules for wireless communication (nrf24l01+, Zigbee Xbee S2C) were also used for obtaining the best performance for this specific purpose and the results have been analyzed in terms of data rate, range and security, and reliability among others. This sensor has been tested on different types of vehicles such as cars, rickshaws and mobile robots such as Husky and Jackal under varying road conditions and the performance was compared with optical encoders that are currently being used for odometry estimation.
Toggle Content
- Must be placed exactly Perpendicular to Ground Plane and Parallel to Wheel Surface. Any vertical angular deviation from this state, results in incorporation of Vehicle’s angular momentum component while turning as noise in the data of wheel angular momentum.
- nRF offers poor reliability, and heavy packet loss.
- Circuit not compact to be used in a small sized bot.
INIOS V.2.0
- New Mathematical Model, to eliminate the constrain of ideal placement.
- Visual Odometry data fused with this sensor to improve accuracy, while not increasing major budget.
- Xbee used to replace nRF.
IMU : Polulu Alt-IMU10-v5
Xbee S2C x3
Arduino Nano
Arduino Mega
Bug-Boost Converter
Charging Module
12V Rechargeable Battery
Undergoing Testing
Undergoing Testing