State-of-the-Art AI Tech Company Spakona Awarded AI Development Project for Next-Generation UAVs
Spakona Inc., in collaboration with Dodwell B.M.S., has been awarded an approximately 500 million JPY project by the Ministry of Defense ATLA to develop a flight demonstration AI for UAVs using multi-agent deep reinforcement learning.
📋 Article Processing Timeline
- 📰 Published: April 7, 2026 at 20:30
- 🔍 Collected: April 7, 2026 at 12:00
- 🤖 AI Analyzed: April 20, 2026 at 22:08 (322h 8m after Collected)
Spakona Inc. (Headquarters: Shibuya-ku, Tokyo; Representative Director: Taro Kawasaki) announces that, in collaboration with Dodwell B.M.S. Ltd. (Headquarters: Chuo-ku, Tokyo; President: Hideyoshi Sasaki, hereinafter "Dodwell B.M.S."), it has been awarded the "Flight Demonstration AI Creation and Flight Test Support Project" promoted by the Aviation Systems Research Center, Acquisition, Technology & Logistics Agency (ATLA).
This project aims at the research and development as well as flight demonstration of Artificial Intelligence (AI) to be equipped on Unmanned Aerial Vehicles (UAVs). Advanced verification using actual aircraft will be conducted over a period of approximately three years.
## Background of the Contract
In recent years, research and development aimed at realizing next-generation systems where manned and unmanned aircraft collaborate has accelerated in the aviation sector. Expectations are rising for AI's advanced decision-making capabilities as the core technology for this. In particular, "multi-agent behavioral decision AI," which makes optimal judgments while considering mutual states in situations where multiple aircraft and systems operate simultaneously, is drawing attention as an important foundational technology supporting next-generation aviation systems.
Under such circumstances, this project aims to establish highly reliable and safe AI by overcoming the challenges that arise when applying models trained in simulation environments to actual aircraft, with an eye toward the practical operation of AI in UAVs. In addition, it tackles advanced technical challenges such as real-time processing with limited computational resources and decision-making under dynamically changing conditions.
These efforts represent a crucial step toward a future where UAVs complement human decision-making and are positioned as one of the advanced research and development projects contributing to the evolution of next-generation aviation technology in Japan.
## Project Overview
In this business, we will develop a "Flight Demonstration AI" that can be mounted on UAVs and verify its performance in both simulation and actual flight environments. The main initiatives are as follows:
### Development of Behavioral Decision AI using Multi-Agent Deep Reinforcement Learning
Assuming an environment where multiple UAVs and virtual agents interact with each other, we will develop an AI utilizing multi-agent deep reinforcement learning. Based on neural networks, the goal is to acquire decision-making capabilities that allow for the selection of optimal actions even in complex situations involving cooperation and competition.
### Implementation of AI Operating in Edge Environments
We will implement an AI model that balances computational efficiency and real-time performance to ensure stable operation on small computers mounted on UAVs.
### Demonstration of Advanced Flight and Tactical Tasks
To verify the behavioral decision-making capabilities in environments involving multiple aircraft, we will conduct advanced tasks such as the following:
- Tracking flight towards virtual targets
- Behavioral decision-making in simulated air-to-air situations by multiple aircraft
- Verification of cooperative and competitive actions among multiple agents
### Verification through HILS Tests and Actual Flight Tests
We will conduct verifications in stages, from ground tests to actual flight tests, to ensure the reliability and safety of the AI in multi-agent environments.
### Continuous AI Improvement and Consideration of Future Prospects
We will continuously improve the AI models based on flight test data and proceed with design considerations looking ahead to their application in future UAV systems.
## Contract Details
Contractor: Aviation Systems Research Center, Acquisition, Technology & Logistics Agency (ATLA)
Scope of Work: Creation of flight demonstration AI for UAVs, and complete flight test support and reporting services
Contract Amount: 503,503,990 JPY
Delivery Date: May 1, 2029
This project aims at the research and development as well as flight demonstration of Artificial Intelligence (AI) to be equipped on Unmanned Aerial Vehicles (UAVs). Advanced verification using actual aircraft will be conducted over a period of approximately three years.
## Background of the Contract
In recent years, research and development aimed at realizing next-generation systems where manned and unmanned aircraft collaborate has accelerated in the aviation sector. Expectations are rising for AI's advanced decision-making capabilities as the core technology for this. In particular, "multi-agent behavioral decision AI," which makes optimal judgments while considering mutual states in situations where multiple aircraft and systems operate simultaneously, is drawing attention as an important foundational technology supporting next-generation aviation systems.
Under such circumstances, this project aims to establish highly reliable and safe AI by overcoming the challenges that arise when applying models trained in simulation environments to actual aircraft, with an eye toward the practical operation of AI in UAVs. In addition, it tackles advanced technical challenges such as real-time processing with limited computational resources and decision-making under dynamically changing conditions.
These efforts represent a crucial step toward a future where UAVs complement human decision-making and are positioned as one of the advanced research and development projects contributing to the evolution of next-generation aviation technology in Japan.
## Project Overview
In this business, we will develop a "Flight Demonstration AI" that can be mounted on UAVs and verify its performance in both simulation and actual flight environments. The main initiatives are as follows:
### Development of Behavioral Decision AI using Multi-Agent Deep Reinforcement Learning
Assuming an environment where multiple UAVs and virtual agents interact with each other, we will develop an AI utilizing multi-agent deep reinforcement learning. Based on neural networks, the goal is to acquire decision-making capabilities that allow for the selection of optimal actions even in complex situations involving cooperation and competition.
### Implementation of AI Operating in Edge Environments
We will implement an AI model that balances computational efficiency and real-time performance to ensure stable operation on small computers mounted on UAVs.
### Demonstration of Advanced Flight and Tactical Tasks
To verify the behavioral decision-making capabilities in environments involving multiple aircraft, we will conduct advanced tasks such as the following:
- Tracking flight towards virtual targets
- Behavioral decision-making in simulated air-to-air situations by multiple aircraft
- Verification of cooperative and competitive actions among multiple agents
### Verification through HILS Tests and Actual Flight Tests
We will conduct verifications in stages, from ground tests to actual flight tests, to ensure the reliability and safety of the AI in multi-agent environments.
### Continuous AI Improvement and Consideration of Future Prospects
We will continuously improve the AI models based on flight test data and proceed with design considerations looking ahead to their application in future UAV systems.
## Contract Details
Contractor: Aviation Systems Research Center, Acquisition, Technology & Logistics Agency (ATLA)
Scope of Work: Creation of flight demonstration AI for UAVs, and complete flight test support and reporting services
Contract Amount: 503,503,990 JPY
Delivery Date: May 1, 2029