[Paper Review] Memory Consistency Conditions for Self-Assembly Programming

Aaron Sterling|ArXiv.org|Sep 15, 2009

Modular Robots and Swarm Intelligence22 references3 citations

TL;DR

This paper establishes a formal connection between self-assembly programming and distributed memory consistency models by reducing the Abstract Tile Assembly Model (aTAM) and Kinetic Tile Assembly Model (kTAM) to distributed shared memory (DSM) systems under causal consistency and GWO, respectively. It shows that local determinism in tile assembly corresponds exactly to concurrent-write freedom in DSM programs, enabling the use of established data-race detection techniques for verifying unique terminal assembly, and demonstrates that self-stabilization theory can be applied to error correction in self-assembly, offering a general framework applicable beyond DNA tiles and 2D surfaces.

ABSTRACT

Perhaps the two most significant theoretical questions about the programming of self-assembling agents are: (1) necessary and sufficient conditions to produce a unique terminal assembly, and (2) error correction. We address both questions, by reducing two well-studied models of tile assembly to models of distributed shared memory (DSM), in order to obtain results from the memory consistency systems induced by tile assembly systems when simulated in the DSM setting. The Abstract Tile Assembly Model (aTAM) can be simulated by a DSM system that obeys causal consistency, and the locally deterministic tile assembly systems in the aTAM correspond exactly to the concurrent-write free programs that simulate tile assembly in such a model. Thus, the detection of the failure of local determinism (which had formerly been an open problem) reduces to the detection of data races in simulating programs. Further, the Kinetic Tile Assembly Model can be simulated by a DSM system that obeys GWO, a memory consistency condition defined by Steinke and Nutt. (To our knowledge, this is the first natural example of a DSM system that obeys GWO, but no stronger consistency condition.) We combine these results with the observation that self-assembly algorithms are local algorithms, and there exists a fast conversion of deterministic local algorithms into deterministic self-stabilizing algorithms. This provides an "immediate" generalization of a theorem by Soloveichik et al. about the existence of tile assembly systems that simultaneously perform two forms of self-stabilization: proofreading and self-healing. Our reductions and proof techniques can be extended to the programming of self-assembling agents in a variety of media, not just DNA tiles, and not just two-dimensional surfaces.

Motivation & Objective

• To address two fundamental theoretical challenges in self-assembly programming: unique terminal assembly and error correction.
• To bridge the gap between self-assembly models and established theories in concurrent programming and distributed systems.
• To generalize results from DNA tile assembly to broader self-assembly contexts, including non-DNA media and higher-dimensional or non-planar structures.
• To demonstrate that self-stabilization techniques can be applied to nanoscale self-assembly for fault tolerance and error correction.
• To identify minimal memory consistency conditions—causal consistency for aTAM and GWO for kTAM—that are sufficient to simulate tile assembly systems.

Proposed method

• Reduced the Abstract Tile Assembly Model (aTAM) to a distributed shared memory (DSM) system under causal consistency, showing that aTAM simulations are equivalent to concurrent-write free programs.
• Mapped locally deterministic tile assembly systems to concurrent-write free programs, enabling data-race detection as a method to verify unique terminal assembly.
• Reduced the Kinetic Tile Assembly Model (kTAM) to a DSM system under the GWO memory consistency condition, which is strictly weaker than causal consistency.
• Used the fact that self-assembly algorithms are local algorithms to apply a known transformation from deterministic local algorithms to deterministic self-stabilizing algorithms.
• Applied this transformation to generalize a result by Soloveichik et al., showing that tile systems can simultaneously perform proofreading and self-healing via self-stabilization.
• Extended the framework to mixed-media self-assembly by modeling tile binding as shared memory operations with specific consistency semantics.

Experimental results

Research questions

• RQ1 What memory consistency condition is sufficient and tight for simulating the Abstract Tile Assembly Model (aTAM) in a distributed shared memory model?
• RQ2 How does local determinism in tile assembly relate to concurrent-write freedom in concurrent programming?
• RQ3 Can the theory of self-stabilization be applied to achieve error correction in kinetic tile assembly systems?
• RQ4 What is the minimal memory consistency condition that captures the behavior of the Kinetic Tile Assembly Model (kTAM)?
• RQ5 How can general-purpose techniques from distributed systems be used to design robust, fault-tolerant self-assembling agents beyond DNA tiles?

Key findings

• The Abstract Tile Assembly Model (aTAM) can be simulated in a distributed shared memory system that obeys causal consistency, and this condition is tight for such simulations.
• Locally deterministic tile assembly systems correspond exactly to concurrent-write free programs in the DSM model, enabling data-race detection to verify unique terminal assembly.
• The Kinetic Tile Assembly Model (kTAM) can be simulated in a DSM system that obeys the GWO memory consistency condition, which is strictly weaker than causal consistency and to our knowledge the first natural example of such a system.
• The reduction to self-stabilizing algorithms generalizes a result by Soloveichik et al., showing that tile systems can simultaneously perform proofreading and self-healing.
• The framework provides a general method for designing self-assembly systems with error correction, applicable to non-DNA media and non-2D surfaces, with a tile complexity upper bound of $|T'| \≤q \binom{4}{3}^2 c^2 |T|^2$ for the simulation of a given tile set $T$.
• The results suggest that weak memory consistency models like GWO can be sufficient for modeling complex self-assembly processes, opening new avenues for fault-tolerant design in nanoscale systems.