mohammad ali sadrnia

This paper mainly aims to determine the optimal drug dosage for the purpose of reducing the population of cancer cells in melanoma patients. To do so, Reinforcement Learning method and the eligibility traces algorithm are employed, giving us the advantage of creating a compromise between the two algorithms of the reinforcement learning, being Monte-Carlo and Temporal Difference. Furthermore, it can be said that using this approach, there was no need to employ a mathematical model in the whole process. However, as its implementation on the real system was not possible, a delayed nonlinear mathematical model is used to investigate the performance of the proposed controller and simulate the behavior of the environment. It should be noted this mathematical model made use of no control method. This is the first time that population control of cancer cells is applied and tested on this model. To know of the optimal dosage of the drug, it should be mentioned that the drug is required to prevent the side effects on healthy/normal cells as much as possible. According to the obtained results, the eligibility traces algorithm is able to control and reduce the population of cancer cells through injecting the sub-optimal drug dose. This will increase the level of immunity in our body. Finally, to demonstrate the advantage of a selective method of increasing the rate of cancer cell death, this method is compared with the Q-learning algorithm and optimal control. By applying the fault to the sensor, the performance of the proposed controller to reduce cancer cells was investigated. The adaptability of the proposed method with the environment changes is checked afterwards. To this end, uncertainty in the system parameters and initial conditions are applied and the population of cancer cells are controlled in five melanoma patients. Moreover, having added noise to the system, it was shown that the eligibility traces algorithm is able to control the population of cancer cells and make it reach zero. Additionally, the convergence speed of both eligibility traces algorithm and Q learning algorithm in reducing the number of cancer cells for different learning rates was investigated.

Melanoma is one of the most dangerous types of cancers and every year, many people suffer from this cancer. Hence, quick diagnosis and treatment are significantly important for the physicians. In the recent decade, intelligent methods have attracted considerable attention for diagnosing and treating the melanoma. The main objective of this paper is determining the optimal dosage of the drug for the elimination of the cancer cells while preventing from the side effect of the drug on the normal cells. To this aim, the Q-learning algorithm is employed. In order to select the actions, a Case-Based Reasoning (CBR) policy is used, which is an accelerated heuristic policy. The considered policy has increased the learning speed and reduced the overall time, to reach the optimal policy. The half-life effect of the drug is also considered to obtain the side effect of the drug on the patientchr('39')s body, at each time step. In order to demonstrate Q-learning algorithm performance in cancer cells control and optimal dosage determination purposes, Q-learning is compared with two methods, including fix dosage injection method and Hamiltonian method, which is one of the most important optimal control methods. Finally, it is revealed that the total injected dosage by using Reinforcement Learning method (Q-learning) is significantly reduced within the whole period of time in comparison with employing the optimal control and a fixed dosage injection cases. The number of cancer cells is controlled, as well. It should be noted that by applying the noise and uncertainty to the system parameters and the initial conditions, the proposed method can successfully control the cancer cells.

In this paper, the main focus is on blood glucose level control and the possible sensor and actuator faults which can be observed in a given system. To this aim, the eligibility traces algorithm (a Reinforcement Learning method) and its combination with sliding mode controllers is used to determine the injection dosage. Through this method, the optimal dosage will be determined to be injected to the patient in order to decrease the side effects of the drug. To detect the fault in the system, residual calculation techniques are utilized. To calculate the residual, it is required to predict states of the normal system at each time step, for which, the Radial Basis Function neural network is used. The proposed method is compared with another reinforcement learning method (Actor-Critic method) with its combination with the sliding mode controller. Finally, both RL-based methods are compared with a combinatory method, Neural network and sliding mode control. Simulation results have revealed that the eligibility traces algorithm and actor-critic method can control the blood glucose concentration and the desired value can be reached, in the presence of the fault. However, in addition to the reduced injected dosage, the eligibility traces algorithm can provide lower variations about the desired value. The reduced injected dosage will result in the mitigated side effects, which will have considerable advantages for diabetic patients.