04-05-2012, 12:41 PM
Real-Time Reactive Control Layer Design- for Intelligent Silver-Mate Robot on RTAI
[attachment=21251]
Introduction :
As medical technology advances with reducing birthrate, many countries are becoming aging or aged societies. To help solve problems of aging society, more and more researches are conducted on development of helper robots namely Silver Robots. These Silver Robots should be capable of handling complex tasks including human-robot interaction via voice and gesture, sensing environments using various sensors, logical inference, as well as performing jobs using manipulation and navigation. In order to perform such complex tasks within given time limits, solid software architecture should be used. Tripodal Schematic Control Architecture (TSCA) developed has been successfully implemented on three versions of their Public Service Robot (PSR) series.
TSCA uses a hybrid approach with three layers for its control: Deliberate layer, Sequencing layer and Reactive layer. The deliberate layer includes software components for human robot interaction and task planning. The sequencing layer has low-level configuration and process supervisor that manages reactive layer components. The
Reactive layer consists of real-time components such as resource components for sensor management, controller components for actuator management, and sets of behavior components for basic real-time control. TSCA of PSR series robots were implemented
on a Windows-based PC, where the reactive layer is not explicit. To achieve real-time performance for robot's sensing, navigation and manipulation, design and implementation software components in the reactive layer using Real-Time Application Interface (RTAI) on Linux. Software components in the reactive layer require real-time performance, since sensing and movement of mobile-base or manipulator have direct effect on the safety issue of Silver-Mate Robots.
Robot Platform:
Hardware Architecture:
Research on real-time reactive control architecture was targeted to a versatile mobile platform. Platform consists of three single board computers (SBC) using P-4M 2.2GHz processor and 1GB SDRAM. Head part consists of two stereo vision cameras, a pan-tilt module, and sixteen microphones for voice processing. vision and voice hardwares are each controlled Separate SBC for application with algorithms requiring heavy computations. All other hardware Components which are largely related to real-time software components are connected to Main SBC as shown in Figure 3. A gyrois connected through data acquisition (DAQ) board. Two laser range finders (LRF), two infra-red (IR) scanners and two sonar controllers, each in charge of six sonar sensors are connected to Main SBC through an 8-port serial board. Two BLDC motors, the only actuators currently on the platform, are controlled by a DSP-based BLDC control board which is also connected through the serial board. BLDC control board also manages encoders attached to each BLDC motors, two front bumper sensors, and also battery voltage.
Software Architecture
Software components comprising the proposed real-time reactive control layer on Main SBC are structured as shown in Figure 4. Each Resource components configures a given sensor hardware for periodic data, acquisition, manages hardware operation, and also stores the sensor data in shared memory for other software components. An Actuator component configures and manages actuator hardware and sends generated control outputs. A Behavior component is a basis of reactive action which uses the sensory data to compute the control output. Finally, a Behavior Coordinator (BC) component collects outputs from a set of Behavior components and fuses them for Actuator component. There are ten Resource components, one for each sensor hardware attached, one Actuator component for the actuator on the platform, one BC component, and two Behavior components. To communicate with components in sequencing layer, RT Task Supervisor (RTTS) component is used to manage all real-time tasks and handle service requests issued to reactive layer.
Real-Time OS and Drivers:
Designed reactive control layer was implemented using RTAI. Version 3.1 of RTAI distribution was used, along with 2.4.27 version of Linux kernel and 0.7.69
Version of COMEDI3 drivers for the DAQ board. Since the Main SBC uses an 8-port serial board, real-time serial driver included in RTAI distribution was modified accordingly. COMEDI driver was also modified to support hardware timer functionality of the DAQ board.
Reactive Control Layer Design
Resource Modules
Gyro Resource:
A gyro sensor is connected to the system through DAQ board. Hence Gyro Resource module uses modified COMEDI driver to convert analog sensor output to digital value periodically, using hardware timer functionality of the DAQ board. Registered Callback function is called when a sensor data is available; hence the callback function stores sensor data in shared memory for other software components.
[attachment=21251]
Introduction :
As medical technology advances with reducing birthrate, many countries are becoming aging or aged societies. To help solve problems of aging society, more and more researches are conducted on development of helper robots namely Silver Robots. These Silver Robots should be capable of handling complex tasks including human-robot interaction via voice and gesture, sensing environments using various sensors, logical inference, as well as performing jobs using manipulation and navigation. In order to perform such complex tasks within given time limits, solid software architecture should be used. Tripodal Schematic Control Architecture (TSCA) developed has been successfully implemented on three versions of their Public Service Robot (PSR) series.
TSCA uses a hybrid approach with three layers for its control: Deliberate layer, Sequencing layer and Reactive layer. The deliberate layer includes software components for human robot interaction and task planning. The sequencing layer has low-level configuration and process supervisor that manages reactive layer components. The
Reactive layer consists of real-time components such as resource components for sensor management, controller components for actuator management, and sets of behavior components for basic real-time control. TSCA of PSR series robots were implemented
on a Windows-based PC, where the reactive layer is not explicit. To achieve real-time performance for robot's sensing, navigation and manipulation, design and implementation software components in the reactive layer using Real-Time Application Interface (RTAI) on Linux. Software components in the reactive layer require real-time performance, since sensing and movement of mobile-base or manipulator have direct effect on the safety issue of Silver-Mate Robots.
Robot Platform:
Hardware Architecture:
Research on real-time reactive control architecture was targeted to a versatile mobile platform. Platform consists of three single board computers (SBC) using P-4M 2.2GHz processor and 1GB SDRAM. Head part consists of two stereo vision cameras, a pan-tilt module, and sixteen microphones for voice processing. vision and voice hardwares are each controlled Separate SBC for application with algorithms requiring heavy computations. All other hardware Components which are largely related to real-time software components are connected to Main SBC as shown in Figure 3. A gyrois connected through data acquisition (DAQ) board. Two laser range finders (LRF), two infra-red (IR) scanners and two sonar controllers, each in charge of six sonar sensors are connected to Main SBC through an 8-port serial board. Two BLDC motors, the only actuators currently on the platform, are controlled by a DSP-based BLDC control board which is also connected through the serial board. BLDC control board also manages encoders attached to each BLDC motors, two front bumper sensors, and also battery voltage.
Software Architecture
Software components comprising the proposed real-time reactive control layer on Main SBC are structured as shown in Figure 4. Each Resource components configures a given sensor hardware for periodic data, acquisition, manages hardware operation, and also stores the sensor data in shared memory for other software components. An Actuator component configures and manages actuator hardware and sends generated control outputs. A Behavior component is a basis of reactive action which uses the sensory data to compute the control output. Finally, a Behavior Coordinator (BC) component collects outputs from a set of Behavior components and fuses them for Actuator component. There are ten Resource components, one for each sensor hardware attached, one Actuator component for the actuator on the platform, one BC component, and two Behavior components. To communicate with components in sequencing layer, RT Task Supervisor (RTTS) component is used to manage all real-time tasks and handle service requests issued to reactive layer.
Real-Time OS and Drivers:
Designed reactive control layer was implemented using RTAI. Version 3.1 of RTAI distribution was used, along with 2.4.27 version of Linux kernel and 0.7.69
Version of COMEDI3 drivers for the DAQ board. Since the Main SBC uses an 8-port serial board, real-time serial driver included in RTAI distribution was modified accordingly. COMEDI driver was also modified to support hardware timer functionality of the DAQ board.
Reactive Control Layer Design
Resource Modules
Gyro Resource:
A gyro sensor is connected to the system through DAQ board. Hence Gyro Resource module uses modified COMEDI driver to convert analog sensor output to digital value periodically, using hardware timer functionality of the DAQ board. Registered Callback function is called when a sensor data is available; hence the callback function stores sensor data in shared memory for other software components.