Marco Platzner
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Marco Platzner has a total of 223 co-authors. Here are the top co-authors by number of shared publications:
- Christian Plessl
Paderborn University, Paderborn, Germany
27 shared publications Paul Kaufmann
Department of Computer Science, Paderborn University, Germany
16 shared publications- Tobias Wiersema
University of Paderborn
8 shared publications Andreas Agne
7 shared publicationsEnno Lubbers
7 shared publications- Heiner Giefers
IBM Research - Zurich, Switzerland
6 shared publications
Latest Publications
- 30 June 2026
- Published by Springer Nature
- p. 34-50
LiveFIFO: FPGA-in-the-Loop Buffer Sizing for Dataflow Accelerators
- 30 June 2026
- Published by Springer Nature
- p. 34-50
LiveFIFO: FPGA-in-the-Loop Buffer Sizing for Dataflow Accelerators
- Open Access
- 4 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (3) , 1-22
Streamlined Dataflow FPGA Architectures for Echo State Neural Networks in Time-Series Prediction and Classification Applications
Echo state networks (ESNs) are reservoir computing (RC) variants that offer comparable accuracy to standard recurrent neural networks at less cost for training, lower computation effort, and higher speed for inference. These characteristics make echo state networks highly suitable for resource-constrained edge implementations. In this article, we present an FPGA-based streamlined dataflow architecture for ESN inference. The accelerator quantizes all layers of the ESN and follows a direct logic implementation style that fully unrolls all computations. We introduce two variants of the accelerator, one that maps neurons to DSP blocks in the FPGA and another one that maps neurons solely to LUTs. We further elaborate on a tool flow to set up, optimize, and train an ESN model for a given dataset and then automatically generate the accelerator designs to be loaded onto an FPGA. We evaluate our accelerators on a number of time-series prediction and classification tasks and compare the errors and accuracies, respectively, for a 32-bit floating-point software baseline and our accelerators at different levels of quantization. We then compare our accelerator with prior work on FPGA-based implementations for ESN and with embedded GPU and CPU platforms. Our experiments show that our accelerators are resource-intensive but excel in latency, throughput, and energy efficiency. For an ESN model with 200 reservoir neurons, we achieve a latency of \(9.5\;\;ns\) , a throughput of \(100\) Megasamples/second, and a power-delay-product of \(72\;\;nWs\) . This outperforms all previous FPGA work and, compared to embedded GPU and CPU platforms, represents improvements in the order of several magnitudes.- 4 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (3) , 1-22
Open AccessStreamlined Dataflow FPGA Architectures for Echo State Neural Networks in Time-Series Prediction and Classification Applications
Echo state networks (ESNs) are reservoir computing (RC) variants that offer comparable accuracy to standard recurrent neural networks at less cost for training, lower computation effort, and higher speed for inference. These characteristics make echo state networks highly suitable for resource-constrained edge implementations. In this article, we present an FPGA-based streamlined dataflow architecture for ESN inference. The accelerator quantizes all layers of the ESN and follows a direct logic implementation style that fully unrolls all computations. We introduce two variants of the accelerator, one that maps neurons to DSP blocks in the FPGA and another one that maps neurons solely to LUTs. We further elaborate on a tool flow to set up, optimize, and train an ESN model for a given dataset and then automatically generate the accelerator designs to be loaded onto an FPGA. We evaluate our accelerators on a number of time-series prediction and classification tasks and compare the errors and accuracies, respectively, for a 32-bit floating-point software baseline and our accelerators at different levels of quantization. We then compare our accelerator with prior work on FPGA-based implementations for ESN and with embedded GPU and CPU platforms. Our experiments show that our accelerators are resource-intensive but excel in latency, throughput, and energy efficiency. For an ESN model with 200 reservoir neurons, we achieve a latency of \(9.5\;\;ns\) , a throughput of \(100\) Megasamples/second, and a power-delay-product of \(72\;\;nWs\) . This outperforms all previous FPGA work and, compared to embedded GPU and CPU platforms, represents improvements in the order of several magnitudes. - Open Access
- 2 September 2025
- Published by MECOnet in WiPiEC Journal - Works in Progress in Embedded Computing Journal
- Vol. 11 (1) , 56-59
CRC: Compressed Reservoir Computing on FPGA via Joint HSIC LASSO-based Pruning and Quantization
While reservoir computing (RC) networks offer advantages over traditional recurrent neural net- works in terms of training time and operational cost for time-series applications, deploying them on edge devices still presents significant challenges due to re- source constraints. Network compression, i.e., pruning and quantization, are thus of utmost importance. We propose a Compressed Reservoir Computing (CRC) framework that integrates advanced pruning and quantization techniques to optimize throughput, latency, energy efficiency, and resource utilization for FPGA- based RC accelerators.We describe the framework with a focus on HSIC LASSO as a novel pruning method that can capture non-linear dependencies between neurons. We validate our framework with time series classification and regression tasks, for which we generate FPGA accelerators. The accelerators achieve a very high throughput of up to 188 Megasamples/s with a latency of 5.32 ns, while reducing resource utilization by 12× and lowering the energy by 10× compared to a baseline hardware implementation, without compromising accuracy.
- 2 September 2025
- Published by MECOnet in WiPiEC Journal - Works in Progress in Embedded Computing Journal
- Vol. 11 (1) , 56-59
Open AccessCRC: Compressed Reservoir Computing on FPGA via Joint HSIC LASSO-based Pruning and Quantization
While reservoir computing (RC) networks offer advantages over traditional recurrent neural net- works in terms of training time and operational cost for time-series applications, deploying them on edge devices still presents significant challenges due to re- source constraints. Network compression, i.e., pruning and quantization, are thus of utmost importance. We propose a Compressed Reservoir Computing (CRC) framework that integrates advanced pruning and quantization techniques to optimize throughput, latency, energy efficiency, and resource utilization for FPGA- based RC accelerators.We describe the framework with a focus on HSIC LASSO as a novel pruning method that can capture non-linear dependencies between neurons. We validate our framework with time series classification and regression tasks, for which we generate FPGA accelerators. The accelerators achieve a very high throughput of up to 188 Megasamples/s with a latency of 5.32 ns, while reducing resource utilization by 12× and lowering the energy by 10× compared to a baseline hardware implementation, without compromising accuracy.
- 2 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (9) , 2395-2405
Design Space Exploration for Approximate Circuits via Checkpointing and DNN-Based Estimators
Approximate computing (AC) is a design paradigm that trades in reductions in hardware area, power, and delay of digital circuits for an increased error. Workflows for synthesizing approximate accelerators are often implemented as search-approximate–evaluate cycles that build a search tree by iteratively approximating components of the original circuit. Design space exploration (DSE) for approximate accelerators is challenging due to the sheer size of the design space and the computational costs for validating the error bound during the search. While early work employed rather slow simulation-based methods, more recent approaches use machine learning (ML) to create fast estimators for validating error bounds. For larger accelerators and design spaces, however, the mispredictions of ML-based methods can steer the search process to explore invalid regions of the design space. In this article, we introduce the framework DeepApprox for approximate accelerator synthesis. The novelty of this framework is its combination of fast ML-based quality estimators with occasional simulations to correct possible mispredictions. The simulations are performed at the so-called checkpoints during the search, and we propose two methods: static and dynamic checkpointing (DC). We present the DeepApprox flow and elaborate on the ML quality estimators and the checkpointing mechanisms. Using a set of smaller and larger benchmarks, we compare DeepApprox with two state-of-the-art flows: a simulation-based method leveraging Monte Carlo tree search (MCTS) and an ML-based method with fast estimators. Our experiments demonstrate that DeepApprox can achieve similar reductions in hardware area and power consumption compared to simulation-based methods, albeit at much lower runtimes, and higher reductions in these metrics compared to methods relying exclusively on ML-based estimators. Thus, by combining ML-based techniques with simulations, DeepApprox provides a scalable DSE for approximate accelerators.
- 2 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (9) , 2395-2405
Design Space Exploration for Approximate Circuits via Checkpointing and DNN-Based Estimators
Approximate computing (AC) is a design paradigm that trades in reductions in hardware area, power, and delay of digital circuits for an increased error. Workflows for synthesizing approximate accelerators are often implemented as search-approximate–evaluate cycles that build a search tree by iteratively approximating components of the original circuit. Design space exploration (DSE) for approximate accelerators is challenging due to the sheer size of the design space and the computational costs for validating the error bound during the search. While early work employed rather slow simulation-based methods, more recent approaches use machine learning (ML) to create fast estimators for validating error bounds. For larger accelerators and design spaces, however, the mispredictions of ML-based methods can steer the search process to explore invalid regions of the design space. In this article, we introduce the framework DeepApprox for approximate accelerator synthesis. The novelty of this framework is its combination of fast ML-based quality estimators with occasional simulations to correct possible mispredictions. The simulations are performed at the so-called checkpoints during the search, and we propose two methods: static and dynamic checkpointing (DC). We present the DeepApprox flow and elaborate on the ML quality estimators and the checkpointing mechanisms. Using a set of smaller and larger benchmarks, we compare DeepApprox with two state-of-the-art flows: a simulation-based method leveraging Monte Carlo tree search (MCTS) and an ML-based method with fast estimators. Our experiments demonstrate that DeepApprox can achieve similar reductions in hardware area and power consumption compared to simulation-based methods, albeit at much lower runtimes, and higher reductions in these metrics compared to methods relying exclusively on ML-based estimators. Thus, by combining ML-based techniques with simulations, DeepApprox provides a scalable DSE for approximate accelerators.
- 4 April 2025
- Published by Springer Nature
- p. 179-195
Ultra-Low Latency and Extreme-Throughput Echo State Neural Networks on FPGA
- 4 April 2025
- Published by Springer Nature
- p. 179-195
Ultra-Low Latency and Extreme-Throughput Echo State Neural Networks on FPGA
- Open Access
- 19 November 2024
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 8 (4) , 244
Correction to: Post-configuration Activation of Hardware Trojans in FPGAs
- 19 November 2024
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 8 (4) , 244
Open AccessCorrection to: Post-configuration Activation of Hardware Trojans in FPGAs
- Open Access
- 13 March 2024
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 8 (2) , 79-93
Post-configuration Activation of Hardware Trojans in FPGAs
The battle of developing hardware Trojans and corresponding countermeasures has taken adversaries towards ingenious ways of compromising hardware designs by circumventing even advanced testing and verification methods. Besides conventional methods of inserting Trojans into a design by a malicious entity, the design flow for field-programmable gate arrays (FPGAs) can also be surreptitiously compromised to perform successful attacks that result in malfunctions or information leakages. In this paper, we introduce a mechanism for the post-configuration activation of a Trojan that leverages malicious routing so that the attacker can leave the Trojan circuit in an undetectable dormant state even in the generated and transmitted bitstream. The Trojan is designed, for example, by adding an enable signal that is routed to an unused primary input/output of the FPGA or by attaching the payload via one route to the remaining design, and then that new route is disconnected during place-and-route and only re-connected when the FPGA is being programmed. The trigger can thus only be activated once the circuit is on the device, which leaves the Trojan dormant in all verification and pre-silicon testing steps. This Trojan can therefore currently neither be prevented by conventional testing and verification methods nor by bitstream-level verification techniques. Since our method ensures that the malicious circuitry is only active in the field, the approach works also quite well with triggerless (always-on) Trojans that have a negligible impact on the overall area and power consumption of the circuit and can thus easily escape detection by fingerprinting techniques using side-channel analyses.
- 13 March 2024
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 8 (2) , 79-93
Open AccessPost-configuration Activation of Hardware Trojans in FPGAs
The battle of developing hardware Trojans and corresponding countermeasures has taken adversaries towards ingenious ways of compromising hardware designs by circumventing even advanced testing and verification methods. Besides conventional methods of inserting Trojans into a design by a malicious entity, the design flow for field-programmable gate arrays (FPGAs) can also be surreptitiously compromised to perform successful attacks that result in malfunctions or information leakages. In this paper, we introduce a mechanism for the post-configuration activation of a Trojan that leverages malicious routing so that the attacker can leave the Trojan circuit in an undetectable dormant state even in the generated and transmitted bitstream. The Trojan is designed, for example, by adding an enable signal that is routed to an unused primary input/output of the FPGA or by attaching the payload via one route to the remaining design, and then that new route is disconnected during place-and-route and only re-connected when the FPGA is being programmed. The trigger can thus only be activated once the circuit is on the device, which leaves the Trojan dormant in all verification and pre-silicon testing steps. This Trojan can therefore currently neither be prevented by conventional testing and verification methods nor by bitstream-level verification techniques. Since our method ensures that the malicious circuitry is only active in the field, the approach works also quite well with triggerless (always-on) Trojans that have a negligible impact on the overall area and power consumption of the circuit and can thus easily escape detection by fingerprinting techniques using side-channel analyses.
- Open Access
- 17 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 23-31
Mapping and Optimizing Communication in ROS 2-based Applications on Configurable System-on-Chip Platforms
The robot operating system is the de-facto standard for designing and implementing robotics applications. Several previous works deal with the integration of heterogeneous accelerators into ROS-based applications. One of these approaches is ReconROS, which enables nodes to be completely mapped to hardware. The follow-up work fpgaDDS extends ReconROS by an intra-FPGA data distribution service to process topic-based communication between nodes entirely in hardware. However, the application of this approach is strictly limited to communication between nodes implemented in hardware only. This paper introduces gateways to close the gap between topic communication in hardware and software. Gateways aim to reduce data transfers between hardware and software by synchronizing a hardware-and software-mapped topic. As a result, data must be transferred only once compared to a separate data transmission for each subscribing hardware node in the baseline. Our measurements show significant speedups in multi-subscriber scenarios with large message sizes. From the conclusions of these measurements, we present a methodology for the communication mapping of ROS 2 computation graphs. In the evaluation, an autonomous driving real-world example benefits from the gateway and achieves a speedup of 1.4.
- 17 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 23-31
Open AccessMapping and Optimizing Communication in ROS 2-based Applications on Configurable System-on-Chip Platforms
The robot operating system is the de-facto standard for designing and implementing robotics applications. Several previous works deal with the integration of heterogeneous accelerators into ROS-based applications. One of these approaches is ReconROS, which enables nodes to be completely mapped to hardware. The follow-up work fpgaDDS extends ReconROS by an intra-FPGA data distribution service to process topic-based communication between nodes entirely in hardware. However, the application of this approach is strictly limited to communication between nodes implemented in hardware only. This paper introduces gateways to close the gap between topic communication in hardware and software. Gateways aim to reduce data transfers between hardware and software by synchronizing a hardware-and software-mapped topic. As a result, data must be transferred only once compared to a separate data transmission for each subscribing hardware node in the baseline. Our measurements show significant speedups in multi-subscriber scenarios with large message sizes. From the conclusions of these measurements, we present a methodology for the communication mapping of ROS 2 computation graphs. In the evaluation, an autonomous driving real-world example benefits from the gateway and achieves a speedup of 1.4.
- Open Access
Mapping and Optimizing Communication in ROS 2-based Applications on Configurable System-on-Chip Platforms
The robot operating system is the de-facto standard for designing and implementing robotics applications. Several previous works deal with the integration of heterogeneous accelerators into ROS-based applications. One of these approaches is ReconROS, which enables nodes to be completely mapped to hardware. The follow-up work fpgaDDS extends ReconROS by an intra-FPGA data distribution service to process topic-based communication between nodes entirely in hardware. However, the application of this approach is strictly limited to communication between nodes implemented in hardware only. This paper introduces gateways to close the gap between topic communication in hardware and software. Gateways aim to reduce data transfers between hardware and software by synchronizing a hardware-and software-mapped topic. As a result, data must be transferred only once compared to a separate data transmission for each subscribing hardware node in the baseline. Our measurements show significant speedups in multi-subscriber scenarios with large message sizes. From the conclusions of these measurements, we present a methodology for the communication mapping of ROS 2 computation graphs. In the evaluation, an autonomous driving real-world example benefits from the gateway and achieves a speedup of 1.4.
Open AccessMapping and Optimizing Communication in ROS 2-based Applications on Configurable System-on-Chip Platforms
The robot operating system is the de-facto standard for designing and implementing robotics applications. Several previous works deal with the integration of heterogeneous accelerators into ROS-based applications. One of these approaches is ReconROS, which enables nodes to be completely mapped to hardware. The follow-up work fpgaDDS extends ReconROS by an intra-FPGA data distribution service to process topic-based communication between nodes entirely in hardware. However, the application of this approach is strictly limited to communication between nodes implemented in hardware only. This paper introduces gateways to close the gap between topic communication in hardware and software. Gateways aim to reduce data transfers between hardware and software by synchronizing a hardware-and software-mapped topic. As a result, data must be transferred only once compared to a separate data transmission for each subscribing hardware node in the baseline. Our measurements show significant speedups in multi-subscriber scenarios with large message sizes. From the conclusions of these measurements, we present a methodology for the communication mapping of ROS 2 computation graphs. In the evaluation, an autonomous driving real-world example benefits from the gateway and achieves a speedup of 1.4.
- Open Access
fpgaDDS: An Intra-FPGA Data Distribution Service for ROS 2 Robotics Applications
Modern computing platforms for robotics applications comprise a set of heterogeneous elements, e.g., multi-core CPUs, embedded GPUs, and FPGAs. FPGAs are reprogrammable hardware devices that allow for fast and energy-efficient computation of many relevant tasks in robotics. ROS is the de-facto programming standard for robotics and decomposes an application into a set of communicating nodes. ReconROS is a previous approach that can map complete ROS nodes into hardware for acceleration. Since ReconROS relies on standard ROS communication layers, exchanging data between hardware-mapped nodes can lead to a performance bottleneck. This paper presents fpgaDDS, a lean data distribution service for hardware-mapped ROS 2 nodes. fpgaDDS relies on a customized and statically generated streaming-based communication architecture. We detail this communication architecture with its components and outline its benefits. We evaluate fpgaDDS on a test example and a larger autonomous vehicle case study. Compared to a ROS 2 application in software, we achieve speedups of up to 13.34 and reduce jitter by two orders of magnitude.
Open AccessfpgaDDS: An Intra-FPGA Data Distribution Service for ROS 2 Robotics Applications
Modern computing platforms for robotics applications comprise a set of heterogeneous elements, e.g., multi-core CPUs, embedded GPUs, and FPGAs. FPGAs are reprogrammable hardware devices that allow for fast and energy-efficient computation of many relevant tasks in robotics. ROS is the de-facto programming standard for robotics and decomposes an application into a set of communicating nodes. ReconROS is a previous approach that can map complete ROS nodes into hardware for acceleration. Since ReconROS relies on standard ROS communication layers, exchanging data between hardware-mapped nodes can lead to a performance bottleneck. This paper presents fpgaDDS, a lean data distribution service for hardware-mapped ROS 2 nodes. fpgaDDS relies on a customized and statically generated streaming-based communication architecture. We detail this communication architecture with its components and outline its benefits. We evaluate fpgaDDS on a test example and a larger autonomous vehicle case study. Compared to a ROS 2 application in software, we achieve speedups of up to 13.34 and reduce jitter by two orders of magnitude.