How rtc-research-group/edip_library works
Main Components
• Preprocessing Modules:
– Background Activity Filter: Implements noise reduction by removing uncorrelated events.
– MaskFilter: Eliminates events from certain pixels based on configuration (e.g., observation period).
• Object Motion Detection and Tracking:
– ObjectMotionDetection and ObjectMotionCell(s): These modules, including the cluster object tracker under folders such as “ClusterObjectTracker” and “Example Project Object Trackers and OMC”, cascade several tracker cells and include velocity estimation and pattern matching.
• Interface / Translation:
– WordSerialDAVIS_to_PAERCAVIAR_translator: Contains the logic to convert the AER format from DAVIS sensors to a different AER format used by another board (PAER-CAVIAR). This serves as a key interfacing block between different hardware standards.
• Motor Control:
– Spike-based PID Motor Controller: Implements a PID controller in the spike-domain using pulse frequency modulation to control DC motors with low latency and power savings.
• ED-BioRob Modules (and related subprojects):
– Contains a wealth of VHDL (and some microcontroller and Java) files targeting different FPGA families (Spartan6, Spartan3) and even an 8051 microcontroller project. These are used for applications such as robot control (BioRob), sensor interfacing, and distributed monitoring.
Relationships and Interactions Between Components
• Data Flow:
– Preprocessing modules (Background Activity Filter and MaskFilter) handle the raw sensor signal first.
– Processed signals then feed into the object motion detection/tracking blocks where cluster object trackers, motion cells, and velocity estimators work in a cascade to detect and analyze movements.
– For scenarios requiring format conversion, the WordSerial-DAVIS to PAER-CAVIAR translator converts data formats so that subsequent modules (or off-chip systems) can understand the signal.
– Finally, the processed data (or controller signals in the case of motion detection) interface with the spike-based PID motor controller to drive actuators.
• Additional Interfaces:
Architectural Patterns and Design Principles
• Modularity and Reusability: Each folder represents a discrete functional block (filtering, tracking, translation, control), illustrating a modular design approach.
• Cascading and Data Pipelining: In the object tracking and motion detection path, several cells (e.g., ObjectMotionCell, ContinuousCounter, etc.) typically operate in cascade, leading to a pipelined processing structure.
• Hardware Interfacing: Notice in ED-BioRob and the translator module there is careful interfacing between different standards and even different device families (FPGA versus microcontroller), emphasizing the importance of clearly defined data buses and combinational/sequential logic boundaries.