Technology
テクノロジー
Our Tech Radar
At Midokura, a subsidiary of Sony Semiconductor Solutions, we are at the forefront of innovation, driving the development of technologies that empower the AITRIOS product line.
Our commitment to democratizing vision-based sensing is underpinned by a suite of cutting-edge technologies. Here, we highlight key technologies that are central to our strategy, each bringing unique benefits and capabilities to our ecosystem.
Check out our latest demo!
Free from low level platform complexities, same app deploy-able on diverse embedded devices with native speed easily.
Our technology enables smart cameras to analyze images directly, reducing data transfer and improving decision-making. We deploy hardware-agnostic Wasm modules containing trained machine learning models, allowing edge devices to make predictions locally.
Additionally, Wasm performs real-time calculations on sensor data streams, facilitating applications like object detection in retail and industry.
Highlighted Topics
1
WebAssembly Containers (Wasm)
WebAssembly containers are the backbone of our application portability strategy. They enable applications to run seamlessly across a wide range of platforms, from tiny microcontrollers (MCUs) to powerful data center processors. This universality is crucial for maintaining consistency, improving efficiency, and fostering a robust ecosystem.
Platform Independence
Write once, run anywhere.
Security
Sandboxed execution environment enhances security.
Efficiency
Low overhead compared to traditional virtual machines.
2
Tiny Edge Computing and MCUs
As IoT devices proliferate, the need for edge computing capabilities on tiny devices, such as 32-bit microcontrollers, becomes paramount. These devices must process data locally to reduce latency, improve responsiveness, and conserve bandwidth.
Reduced Latency
Local processing speeds up decision-making.
Bandwidth Efficiency
Less data transmission to the cloud.
Energy Efficiency
Lower power consumption suited for battery-operated devices.
3
IoT Protocols (MQTT, CoAp)
Efficient communication protocols are essential for IoT devices to interact with each other and with central systems. MQTT and CoAp are lightweight protocols designed for the constraints of IoT environments, ensuring reliable and efficient data exchange.
Lightweight
Minimal overhead suited for constrained devices.
Reliability
Quality of service levels for assured message delivery.
Scalability
Efficiently handles numerous devices and connections.
4
RISC-V
RISC-V is an open standard instruction set architecture (ISA) that provides a flexible and extensible foundation for building custom processors. Its openness promotes innovation and enables customization for specific application needs.
Open Standard
No licensing fees, fostering innovation.
Customizable
Tailor the ISA to specific requirements.
Ecosystem Growth
Rapidly growing community and industry support.
5
OpenAMP
OpenAMP (Open Asymmetric Multi-Processing) facilitates communication and resource sharing between heterogeneous processors, which is essential for complex IoT devices that leverage multiple types of processors for different tasks.
Inter-Processor Communication
Seamless data exchange between different processors.
Resource Sharing
Efficient utilization of system resources.
Scalability
Supports complex, multi-core systems.
6
IoT Protocols (MQTT, CoAp)
TinyML brings the power of machine learning to constrained devices, enabling them to perform intelligent tasks locally. This technology is critical for applications requiring real-time decision-making and minimal latency.
Local Intelligence
ML models running on-device for immediate insights.
Low Latency
Instantaneous processing and response.
Energy Efficiency
Optimized for low-power consumption.
7
Polyglot SDKs
Supporting multiple programming languages (polyglot development) is crucial for attracting a diverse developer base. Our SDKs enable developers to use their preferred languages, increasing productivity and innovation.
Flexibility
Use the best language for the task.
Developer Satisfaction
Attracts and retains diverse talent.
Innovation
Encourages experimentation with different languages and tools.
8
OpenAMP Debugging Tools
Effective debugging tools are essential for identifying and fixing issues quickly. Our suite of debugging tools ensures developers can diagnose problems efficiently, reducing development time and improving product quality.
Efficiency
Rapid identification and resolution of bugs.
Ease of Use
Intuitive interfaces and comprehensive features.
Product Quality
Enhanced reliability and performance.
9
Performance/Profiling Tools
To ensure applications run optimally across various devices, performance and profiling tools are necessary. These tools help developers understand resource usage, identify bottlenecks, and optimize code.
Optimization
Improved application performance.
Resource Management
Efficient use of hardware capabilities.
Insight
Detailed analysis of code execution.
10
Observability Tools
Observability tools provide deep insights into system behavior and performance, enabling proactive monitoring and troubleshooting. They are crucial for maintaining high availability and reliability in complex IoT systems.
Real-Time Monitoring
Immediate insights into system status.
Proactive Troubleshooting
Identify and address issues before they impact users.
Reliability
Ensure consistent and reliable operation of IoT devices.
Discover our technology!
We work with cutting edge technology to create the solutions of the future
Vision Sensing Application Creation
- Creation of tools for easy Vision Sensing Application generation for the edge.
- Use your trained models for creating applications powered with AI.
- Build the block of your application with WebAssembly (Wasm) modules and connect them.
- Compilation ahead of time for modules and model optimization for target devices.
Vision Sensing Application Lifecycle Management
- Help developers to manage the application lifecycle.
- Distribute the logic among multiple devices, offloading compute from weak devices to others with more computational power.
- Abstract away the differences between different service providers and IoT platforms.
- Provide a consistent set of services (Security, distributed storage, communication…) to managed applications.
Vision Sensing Application Execution and Monitoring in the Edge
- We work on fail-safe deployment of secure, sandboxed components without requiring OTA updates.
- Targeting edge devices ranging from larger Cortex-A microprocessors to resource-constrained microcontrollers.
- Allowing user defined component pipelines to be securely deployed to IoT device fleets.
AI Model Training and Optimization as a Service
- Separate the concepts of Model creation, Domain adaptation and Fine Tuning to ease Model creation and allow cross-company collaboration.
- Train an already existing Model just by adding your images and annotations, or create your own Model from scratch.
- Ensure your trained model is compatible with target devices in advance.
AI Models Creation and Research
- Research new network architectures in TensorFlow and Pytorch fitting into Edge Devices and Smart Cameras.
- Research training methods and model optimization like Transfer Learning and Knowledge Distillation.
- Development of an SDK for integrating AI models into the AITRIOS life cycle.
