Lightweight Evolution Strategies for Nanoswimmers-oriented In Vivo Computation

Shi, Shaolong; Chen, Yifan; Yao, Xin; Zhang, Mengjie

doi:10.1109/cec.2019.8790356

Cited by 12 publications

(7 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thirdly, the fitness evaluation of computing agents is dependent on the estimation of the BGF landscape by analyzing the measurable characteristics of nanorobots, which is different from the fitness evaluation in traditional optimization without any requirement for on-the-fly estimation [20]. Next, the nanorobots are controlled by the external actuating system with a uniform magnetic field, which means that the other nanorobots will be influenced when one is actuated towards its intended direction [28], [29]. Furthermore, constrained by the control accuracy and imaging resolution of the external devices, the steering and positioning errors in each iteration should be considered [30].…”

Section: A Tumor Sensitization and Targeting As In Vivo Computationmentioning

confidence: 99%

“…IA/UA is used to actuate the nanorobots by the external actuating system while IT aims at exploiting the BGF estimated from the nanorobot characteristics measured by an external tracking system. To realize the optimal control of nanorobots in the IA/UA mode with the homogeneous magnetic field generated by the state-of-the-art actuating device, we have proposed a novel evolution strategy, namely the weak priority evolution strategy (WP-ES), which has been used to find an optimal movement direction of nanorobots in each iteration of the swarm intelligence algorithm [29].…”

Section: B Review Of Previous Workmentioning

confidence: 99%

“…In the TST process, an external actuating device maneuvers nanorobot swarms to find the tumor lesions according to a specific natural computational algorithm. The tracking system is used to locate the nanorobot swarms and estimate the BGFs according to the measurable characteristics of nanorobots, which are used for the fitness evaluation for the computing agents [19], [29]. In this way, the nanorobots update their locations step by step until the maximum search time is reached.…”

Section: A Problem Formulationmentioning

confidence: 99%

“…where U is an apparent objective function representing BGF, which is measured by computing agent A; x ∈ S is a bivector representing the possible location of any agent with S being the search space; x * represent the locations of the global optimal solutions that remain unchanged regardless of any variation caused by the agents to the BGF landscape. Furthermore, the apparent objective function contains the true objective function U T (x) independent of the presence or absence of agent A, the disturbance caused by the agent A to the search space S, and the correction factor accounting for this disturbance as described in [29]. Thus, with the correction factor attempting to counteract the disturbance to minimize its influence on the true BGF landscape, the apparent objective function is as follows:…”

Section: A Problem Formulationmentioning

confidence: 99%

“…To evaluate the fitness of nanorobots which are natural computing agents during the targeting process, we introduce the concept of BGF, which is derived from the biological characteristics of high-risk tissue regions [29], [33]. For example, the healthy capillaries exhibit almost uniformly distributed grid patterns while the tumor vessels have a profound sort of tortuosity with the density increasing around the lesions; the locations of nanorobots can be estimated through contrastenhanced medical imaging according to the magnetic response induced by a polarizing magnetic field [42], [43].…”

Section: Representative Bgf Landscapesmentioning

confidence: 99%

See 4 more Smart Citations

Nanorobots-Assisted Natural Computation for Multifocal Tumor Sensitization and Targeting

Shi

Yan

Xiong

et al. 2021

IEEE Trans.on Nanobioscience

Self Cite

View full text Add to dashboard Cite

We have proposed a new tumor sensitization and targeting (TST) framework, named in vivo computation, in our previous investigations. The problem of TST for an early and microscopic tumor is interpreted from the computational perspective with nanorobots being the "natural" computing agents, the high-risk tissue being the search space, the tumor targeted being the global optimal solution, and the tumortriggered biological gradient field (BGF) providing the aided knowledge for fitness evaluation of nanorobots. This natural computation process can be seen as on-the-fly path planning for nanorobot swarms with an unknown target position, which is different from the traditional path planning methods. Our previous works are focusing on the TST for a solitary lesion, where we proposed the weak priority evolution strategy (WP-ES) to adapt to the actuating mode of the homogeneous magnetic field used in the state-of-the-art nanorobotic platforms, and some in vitro validations were performed. In this paper, we focus on the problem of TST for multifocal tumors, which can be seen as a multimodal optimization problem for the "natural" computation. To overcome this issue, we propose a sequential targeting strategy (Se-TS) to complete TST for the multiple lesions with the assistance of nanorobot swarms, which are maneuvered by the external actuating and tracking devices according to the WP-ES. The Se-TS is used to modify the BGF landscape after a tumor is detected by a nanorobot swarm with the gathered BGF information around the detected tumor. Next, another nanorobot swarm will be employed to find the second tumor according to the modified BGF landscape without being misguided to the previous one. In this way, all the tumor lesions will be detected one by one. In other words, the paths of nanorobots to find the targets can be generated successively with the sequential modification of the BGF landscape. To demonstrate the effectiveness of the proposed Se-TS, we perform comprehensive simulation studies by enhancing the WP-ES based swarm intelligence algorithms using this strategy considering the realistic in-body constraints. The performance is compared against that of the "brute-force" search, which corresponds to the traditional systemic tumor targeting, and also against that of the standard swarm intelligence algorithms from the algorithmic perspective. Furthermore, some in vitro experiments are performed by using Janus microparticles as

show abstract