Periodic Reporting for period 1 - GUIDO (A Computing Unit for Autonomous Spacecraft Guidance in Deep Space)
Reporting period: 2023-06-01 to 2024-11-30
The “space revolution” is expected to become the biggest industry in history. During the last decades, space exploration has contributed to the development of various technologies (e.g. GPS, weather predictions, solar cells, reliable power grid, Earth observation services) that fed-back into the economy and improved our lives. In 2020, the global space economy was valued at about $ 446.9 B. With the advent of interplanetary CubeSats, nanosatellites that disrupted this industry, space agencies are planning to launch tens of them in deep space per single launch.
The number of deep-space probes is bound to thrive, yet their growth is unsustainable with current practice. Human-in-the-loop operations are particularly relevant for flight-related procedures in cruise phase: the spacecraft position is determined from ground, with a consistent involvement of resources and assets. It is acknowledged that a sustainable deep-space exploration requires enabling self-driving spacecraft.
The GUIDO project forges an autonomous guidance board to enable the self-driving capability of spacecraft in the solar system. The project stems from the ERC-funded project EXTREMA (ERC-CoG-2019). GUIDO aims to elevate the convex-optimization-based guidance algorithms conceived within EXTREMA to a proof-of-concept board at TRL 4.
Our autonomous guidance board, GUIDOboard, is a disruptive innovation that will change the modus operandi of deep-space satellite operation. GUIDO will revolutionize the way spacecraft travel in space, by cutting the cost of ground control of the whole cruise phase and unlocking the new space mission concept.
The project follows a suitably devised pathway to cover technical validation activities and assessment of exploitation opportunities. GUIDO has raised the interest of leading sectoral players and space agencies, as witnessed by the letters of interest attached.
The number of deep-space probes is bound to thrive, yet their growth is unsustainable with current practice. Human-in-the-loop operations are particularly relevant for flight-related procedures in cruise phase: the spacecraft position is determined from ground, with a consistent involvement of resources and assets. It is acknowledged that a sustainable deep-space exploration requires enabling self-driving spacecraft.
The GUIDO project forges an autonomous guidance board to enable the self-driving capability of spacecraft in the solar system. The project stems from the ERC-funded project EXTREMA (ERC-CoG-2019). GUIDO aims to elevate the convex-optimization-based guidance algorithms conceived within EXTREMA to a proof-of-concept board at TRL 4.
Our autonomous guidance board, GUIDOboard, is a disruptive innovation that will change the modus operandi of deep-space satellite operation. GUIDO will revolutionize the way spacecraft travel in space, by cutting the cost of ground control of the whole cruise phase and unlocking the new space mission concept.
The project follows a suitably devised pathway to cover technical validation activities and assessment of exploitation opportunities. GUIDO has raised the interest of leading sectoral players and space agencies, as witnessed by the letters of interest attached.
The GUIDO project is articulated in three Actions (see the figure "Overview of the project activities" attached).
Action 1: Product Prototyping and Optimization
This activity included the delta-developments required for the deployment and optimization of the guidance algorithm on the target hardware. First, a software and processor in the loop profiling has been carried out to identify the most promising function which could be optimized, both in terms of number of calls and computational time. A study on the memory allocation was then performed to understand possible area of improvement of the convex solver. A trade off analysis with different development kits was performed, identifying the Kria KD240 as the target for the deployment, given its small power envelope and computational capabilities. An image of the board integrated with the ETHILE facility is attached (see the figure "Kria board deployed and interfaces with the ETHILE facility").
First, the inputs and outputs file formats for the guidance algorithm have been standardized to allow for test repeatability, debugging, and cope with on-board execution requirements. Protobufs were exploited to unify and simplify management of messages between facilities employing different programming languages. Then, the algorithm deployment procedure was consolidated and a set of unit and regression tests introduced to monitor the effectiveness of the deployment and identify undesired modifications. Finally, a communication layer with the UDP protocol was developed and tested to allow for the commanding the computed solutions with the ETHILE facility.
Action 2: Validation – Autonomous Guidance Unit Validation
As part of the validation activities, some delta-development identified during the first batch of simulations have been performed to improve the simulation realism and overcome some minor issues identified during the execution of the first set of simulations. These include the time synchronization between the ETHILE facility and SPESI and the deployment of the optimized algorithm on the Kria. Updates to the calibration procedure to account for the new setup were introduced. Intermediate step-by-step subsystem testing were performed to deal with the additional complexity introduced by the presence of the hardware. The sensor validation has been performed on different transfer scenarios. Two configurations were studied: one with the algorithm running fully on the processing system and a hybrid configuration exploiting both the processing system and the programmable logic on the Kria. The tests were performed in open-loop, and the ETHILE facility was used to simulate the actual thrust noise coming from the thruster actuation. The orbit propagation is performed by SPESI. Simulations showed that the power required is below 3W, with large margins in terms of programmable logic resources and algorithm (solver) optimization.
Action 3 – Exploitation and Knowledge Transfer: Market Assessment and Exploitation
This activity combined desk research aimed at outlining the knowledge landscape, the structure of the supply chain, and the market outlook of the small satellite domain and exercises/brainstorming sessions aimed at further reflecting on the value proposition of the GUIDOboard and its competitive advantages, inputs for re-examining the business proposition. For what concerns the market analysis, the outcomes were an overview of the guidance computers sector and market (its profile, dimension, and trend) and competitive intelligence analysis.
Action 1: Product Prototyping and Optimization
This activity included the delta-developments required for the deployment and optimization of the guidance algorithm on the target hardware. First, a software and processor in the loop profiling has been carried out to identify the most promising function which could be optimized, both in terms of number of calls and computational time. A study on the memory allocation was then performed to understand possible area of improvement of the convex solver. A trade off analysis with different development kits was performed, identifying the Kria KD240 as the target for the deployment, given its small power envelope and computational capabilities. An image of the board integrated with the ETHILE facility is attached (see the figure "Kria board deployed and interfaces with the ETHILE facility").
First, the inputs and outputs file formats for the guidance algorithm have been standardized to allow for test repeatability, debugging, and cope with on-board execution requirements. Protobufs were exploited to unify and simplify management of messages between facilities employing different programming languages. Then, the algorithm deployment procedure was consolidated and a set of unit and regression tests introduced to monitor the effectiveness of the deployment and identify undesired modifications. Finally, a communication layer with the UDP protocol was developed and tested to allow for the commanding the computed solutions with the ETHILE facility.
Action 2: Validation – Autonomous Guidance Unit Validation
As part of the validation activities, some delta-development identified during the first batch of simulations have been performed to improve the simulation realism and overcome some minor issues identified during the execution of the first set of simulations. These include the time synchronization between the ETHILE facility and SPESI and the deployment of the optimized algorithm on the Kria. Updates to the calibration procedure to account for the new setup were introduced. Intermediate step-by-step subsystem testing were performed to deal with the additional complexity introduced by the presence of the hardware. The sensor validation has been performed on different transfer scenarios. Two configurations were studied: one with the algorithm running fully on the processing system and a hybrid configuration exploiting both the processing system and the programmable logic on the Kria. The tests were performed in open-loop, and the ETHILE facility was used to simulate the actual thrust noise coming from the thruster actuation. The orbit propagation is performed by SPESI. Simulations showed that the power required is below 3W, with large margins in terms of programmable logic resources and algorithm (solver) optimization.
Action 3 – Exploitation and Knowledge Transfer: Market Assessment and Exploitation
This activity combined desk research aimed at outlining the knowledge landscape, the structure of the supply chain, and the market outlook of the small satellite domain and exercises/brainstorming sessions aimed at further reflecting on the value proposition of the GUIDOboard and its competitive advantages, inputs for re-examining the business proposition. For what concerns the market analysis, the outcomes were an overview of the guidance computers sector and market (its profile, dimension, and trend) and competitive intelligence analysis.
The main outcomes of Action 1 are:
- Identification of a target board for guidance algorithm execution.
- Integration of the target development board within the EXTREMA Simulation Hub.
Whereas, the main outcomes of Action 2 are:
- Validation of the guidance algorithm on a computing unit using both a Processing System and a Programmable Logic with minimal power consumption.
- Validation of the guidance algorithm with thruster facility in the loop in an accelerated framework.
For what concerns both Actions 1 and 2, further optimizations and improvements are possible since there is still room for improvement both in terms of board capabilities and algorithm optimization for execution of FPGAs. The simulations executed were only in open loop, with a full trajectory executions and no re-computations after a thrust arc. Full simulations involving also the attitude commanding and the other components of the EXTREMA simulation hub are required for a complete validation of the supervision logic, which shall monitor and control the execution of the guidance.
Action 3 yielded the following main outcomes:
- Exploration of the "Global Deep Space Guidance Computers and Planning Software" market for the identification of opportunities trends, and competitors.
- Formulation of an exploitation plan, comprehending the development of a value proposition, identification of funding schemes for future advancement, and an examination of the pros and cons associated with exploitation routes (licensing and direct use).
The key needs to ensure further uptakes are:
- Looking for additional funding to increase the sensor TRL
- Establishing a collaborative partnership with a company responsible for the development of the guidance unit hardware, starting from the same architecture of the Kria.
- Research possible opportunities that can expand the guidance unit and guidance software capabilities to address in a shorter timeframe a larger SOM.
- Identification of a target board for guidance algorithm execution.
- Integration of the target development board within the EXTREMA Simulation Hub.
Whereas, the main outcomes of Action 2 are:
- Validation of the guidance algorithm on a computing unit using both a Processing System and a Programmable Logic with minimal power consumption.
- Validation of the guidance algorithm with thruster facility in the loop in an accelerated framework.
For what concerns both Actions 1 and 2, further optimizations and improvements are possible since there is still room for improvement both in terms of board capabilities and algorithm optimization for execution of FPGAs. The simulations executed were only in open loop, with a full trajectory executions and no re-computations after a thrust arc. Full simulations involving also the attitude commanding and the other components of the EXTREMA simulation hub are required for a complete validation of the supervision logic, which shall monitor and control the execution of the guidance.
Action 3 yielded the following main outcomes:
- Exploration of the "Global Deep Space Guidance Computers and Planning Software" market for the identification of opportunities trends, and competitors.
- Formulation of an exploitation plan, comprehending the development of a value proposition, identification of funding schemes for future advancement, and an examination of the pros and cons associated with exploitation routes (licensing and direct use).
The key needs to ensure further uptakes are:
- Looking for additional funding to increase the sensor TRL
- Establishing a collaborative partnership with a company responsible for the development of the guidance unit hardware, starting from the same architecture of the Kria.
- Research possible opportunities that can expand the guidance unit and guidance software capabilities to address in a shorter timeframe a larger SOM.